1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 15 * Bjorn Ekwall. <bj0rn@blox.se> 16 * Pekka Riikonen <priikone@poseidon.pspt.fi> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/dmaengine.h> 39 #include <linux/workqueue.h> 40 #include <linux/dynamic_queue_limits.h> 41 42 #include <linux/ethtool.h> 43 #include <net/net_namespace.h> 44 #include <net/dsa.h> 45 #ifdef CONFIG_DCB 46 #include <net/dcbnl.h> 47 #endif 48 #include <net/netprio_cgroup.h> 49 50 #include <linux/netdev_features.h> 51 #include <linux/neighbour.h> 52 #include <uapi/linux/netdevice.h> 53 #include <uapi/linux/if_bonding.h> 54 55 struct netpoll_info; 56 struct device; 57 struct phy_device; 58 /* 802.11 specific */ 59 struct wireless_dev; 60 /* 802.15.4 specific */ 61 struct wpan_dev; 62 struct mpls_dev; 63 64 void netdev_set_default_ethtool_ops(struct net_device *dev, 65 const struct ethtool_ops *ops); 66 67 /* Backlog congestion levels */ 68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 69 #define NET_RX_DROP 1 /* packet dropped */ 70 71 /* 72 * Transmit return codes: transmit return codes originate from three different 73 * namespaces: 74 * 75 * - qdisc return codes 76 * - driver transmit return codes 77 * - errno values 78 * 79 * Drivers are allowed to return any one of those in their hard_start_xmit() 80 * function. Real network devices commonly used with qdiscs should only return 81 * the driver transmit return codes though - when qdiscs are used, the actual 82 * transmission happens asynchronously, so the value is not propagated to 83 * higher layers. Virtual network devices transmit synchronously, in this case 84 * the driver transmit return codes are consumed by dev_queue_xmit(), all 85 * others are propagated to higher layers. 86 */ 87 88 /* qdisc ->enqueue() return codes. */ 89 #define NET_XMIT_SUCCESS 0x00 90 #define NET_XMIT_DROP 0x01 /* skb dropped */ 91 #define NET_XMIT_CN 0x02 /* congestion notification */ 92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 94 95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 96 * indicates that the device will soon be dropping packets, or already drops 97 * some packets of the same priority; prompting us to send less aggressively. */ 98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 100 101 /* Driver transmit return codes */ 102 #define NETDEV_TX_MASK 0xf0 103 104 enum netdev_tx { 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 106 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 109 }; 110 typedef enum netdev_tx netdev_tx_t; 111 112 /* 113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 115 */ 116 static inline bool dev_xmit_complete(int rc) 117 { 118 /* 119 * Positive cases with an skb consumed by a driver: 120 * - successful transmission (rc == NETDEV_TX_OK) 121 * - error while transmitting (rc < 0) 122 * - error while queueing to a different device (rc & NET_XMIT_MASK) 123 */ 124 if (likely(rc < NET_XMIT_MASK)) 125 return true; 126 127 return false; 128 } 129 130 /* 131 * Compute the worst case header length according to the protocols 132 * used. 133 */ 134 135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 136 # if defined(CONFIG_MAC80211_MESH) 137 # define LL_MAX_HEADER 128 138 # else 139 # define LL_MAX_HEADER 96 140 # endif 141 #else 142 # define LL_MAX_HEADER 32 143 #endif 144 145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 147 #define MAX_HEADER LL_MAX_HEADER 148 #else 149 #define MAX_HEADER (LL_MAX_HEADER + 48) 150 #endif 151 152 /* 153 * Old network device statistics. Fields are native words 154 * (unsigned long) so they can be read and written atomically. 155 */ 156 157 struct net_device_stats { 158 unsigned long rx_packets; 159 unsigned long tx_packets; 160 unsigned long rx_bytes; 161 unsigned long tx_bytes; 162 unsigned long rx_errors; 163 unsigned long tx_errors; 164 unsigned long rx_dropped; 165 unsigned long tx_dropped; 166 unsigned long multicast; 167 unsigned long collisions; 168 unsigned long rx_length_errors; 169 unsigned long rx_over_errors; 170 unsigned long rx_crc_errors; 171 unsigned long rx_frame_errors; 172 unsigned long rx_fifo_errors; 173 unsigned long rx_missed_errors; 174 unsigned long tx_aborted_errors; 175 unsigned long tx_carrier_errors; 176 unsigned long tx_fifo_errors; 177 unsigned long tx_heartbeat_errors; 178 unsigned long tx_window_errors; 179 unsigned long rx_compressed; 180 unsigned long tx_compressed; 181 }; 182 183 184 #include <linux/cache.h> 185 #include <linux/skbuff.h> 186 187 #ifdef CONFIG_RPS 188 #include <linux/static_key.h> 189 extern struct static_key rps_needed; 190 #endif 191 192 struct neighbour; 193 struct neigh_parms; 194 struct sk_buff; 195 196 struct netdev_hw_addr { 197 struct list_head list; 198 unsigned char addr[MAX_ADDR_LEN]; 199 unsigned char type; 200 #define NETDEV_HW_ADDR_T_LAN 1 201 #define NETDEV_HW_ADDR_T_SAN 2 202 #define NETDEV_HW_ADDR_T_SLAVE 3 203 #define NETDEV_HW_ADDR_T_UNICAST 4 204 #define NETDEV_HW_ADDR_T_MULTICAST 5 205 bool global_use; 206 int sync_cnt; 207 int refcount; 208 int synced; 209 struct rcu_head rcu_head; 210 }; 211 212 struct netdev_hw_addr_list { 213 struct list_head list; 214 int count; 215 }; 216 217 #define netdev_hw_addr_list_count(l) ((l)->count) 218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 219 #define netdev_hw_addr_list_for_each(ha, l) \ 220 list_for_each_entry(ha, &(l)->list, list) 221 222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 224 #define netdev_for_each_uc_addr(ha, dev) \ 225 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 226 227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 229 #define netdev_for_each_mc_addr(ha, dev) \ 230 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 231 232 struct hh_cache { 233 u16 hh_len; 234 u16 __pad; 235 seqlock_t hh_lock; 236 237 /* cached hardware header; allow for machine alignment needs. */ 238 #define HH_DATA_MOD 16 239 #define HH_DATA_OFF(__len) \ 240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 241 #define HH_DATA_ALIGN(__len) \ 242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 244 }; 245 246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 247 * Alternative is: 248 * dev->hard_header_len ? (dev->hard_header_len + 249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 250 * 251 * We could use other alignment values, but we must maintain the 252 * relationship HH alignment <= LL alignment. 253 */ 254 #define LL_RESERVED_SPACE(dev) \ 255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 258 259 struct header_ops { 260 int (*create) (struct sk_buff *skb, struct net_device *dev, 261 unsigned short type, const void *daddr, 262 const void *saddr, unsigned int len); 263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 265 void (*cache_update)(struct hh_cache *hh, 266 const struct net_device *dev, 267 const unsigned char *haddr); 268 }; 269 270 /* These flag bits are private to the generic network queueing 271 * layer, they may not be explicitly referenced by any other 272 * code. 273 */ 274 275 enum netdev_state_t { 276 __LINK_STATE_START, 277 __LINK_STATE_PRESENT, 278 __LINK_STATE_NOCARRIER, 279 __LINK_STATE_LINKWATCH_PENDING, 280 __LINK_STATE_DORMANT, 281 }; 282 283 284 /* 285 * This structure holds at boot time configured netdevice settings. They 286 * are then used in the device probing. 287 */ 288 struct netdev_boot_setup { 289 char name[IFNAMSIZ]; 290 struct ifmap map; 291 }; 292 #define NETDEV_BOOT_SETUP_MAX 8 293 294 int __init netdev_boot_setup(char *str); 295 296 /* 297 * Structure for NAPI scheduling similar to tasklet but with weighting 298 */ 299 struct napi_struct { 300 /* The poll_list must only be managed by the entity which 301 * changes the state of the NAPI_STATE_SCHED bit. This means 302 * whoever atomically sets that bit can add this napi_struct 303 * to the per-cpu poll_list, and whoever clears that bit 304 * can remove from the list right before clearing the bit. 305 */ 306 struct list_head poll_list; 307 308 unsigned long state; 309 int weight; 310 unsigned int gro_count; 311 int (*poll)(struct napi_struct *, int); 312 #ifdef CONFIG_NETPOLL 313 spinlock_t poll_lock; 314 int poll_owner; 315 #endif 316 struct net_device *dev; 317 struct sk_buff *gro_list; 318 struct sk_buff *skb; 319 struct hrtimer timer; 320 struct list_head dev_list; 321 struct hlist_node napi_hash_node; 322 unsigned int napi_id; 323 }; 324 325 enum { 326 NAPI_STATE_SCHED, /* Poll is scheduled */ 327 NAPI_STATE_DISABLE, /* Disable pending */ 328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 329 NAPI_STATE_HASHED, /* In NAPI hash */ 330 }; 331 332 enum gro_result { 333 GRO_MERGED, 334 GRO_MERGED_FREE, 335 GRO_HELD, 336 GRO_NORMAL, 337 GRO_DROP, 338 }; 339 typedef enum gro_result gro_result_t; 340 341 /* 342 * enum rx_handler_result - Possible return values for rx_handlers. 343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 344 * further. 345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 346 * case skb->dev was changed by rx_handler. 347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 349 * 350 * rx_handlers are functions called from inside __netif_receive_skb(), to do 351 * special processing of the skb, prior to delivery to protocol handlers. 352 * 353 * Currently, a net_device can only have a single rx_handler registered. Trying 354 * to register a second rx_handler will return -EBUSY. 355 * 356 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 357 * To unregister a rx_handler on a net_device, use 358 * netdev_rx_handler_unregister(). 359 * 360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 361 * do with the skb. 362 * 363 * If the rx_handler consumed to skb in some way, it should return 364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 365 * the skb to be delivered in some other ways. 366 * 367 * If the rx_handler changed skb->dev, to divert the skb to another 368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 369 * new device will be called if it exists. 370 * 371 * If the rx_handler consider the skb should be ignored, it should return 372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 373 * are registered on exact device (ptype->dev == skb->dev). 374 * 375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 376 * delivered, it should return RX_HANDLER_PASS. 377 * 378 * A device without a registered rx_handler will behave as if rx_handler 379 * returned RX_HANDLER_PASS. 380 */ 381 382 enum rx_handler_result { 383 RX_HANDLER_CONSUMED, 384 RX_HANDLER_ANOTHER, 385 RX_HANDLER_EXACT, 386 RX_HANDLER_PASS, 387 }; 388 typedef enum rx_handler_result rx_handler_result_t; 389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 390 391 void __napi_schedule(struct napi_struct *n); 392 void __napi_schedule_irqoff(struct napi_struct *n); 393 394 static inline bool napi_disable_pending(struct napi_struct *n) 395 { 396 return test_bit(NAPI_STATE_DISABLE, &n->state); 397 } 398 399 /** 400 * napi_schedule_prep - check if napi can be scheduled 401 * @n: napi context 402 * 403 * Test if NAPI routine is already running, and if not mark 404 * it as running. This is used as a condition variable 405 * insure only one NAPI poll instance runs. We also make 406 * sure there is no pending NAPI disable. 407 */ 408 static inline bool napi_schedule_prep(struct napi_struct *n) 409 { 410 return !napi_disable_pending(n) && 411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 412 } 413 414 /** 415 * napi_schedule - schedule NAPI poll 416 * @n: napi context 417 * 418 * Schedule NAPI poll routine to be called if it is not already 419 * running. 420 */ 421 static inline void napi_schedule(struct napi_struct *n) 422 { 423 if (napi_schedule_prep(n)) 424 __napi_schedule(n); 425 } 426 427 /** 428 * napi_schedule_irqoff - schedule NAPI poll 429 * @n: napi context 430 * 431 * Variant of napi_schedule(), assuming hard irqs are masked. 432 */ 433 static inline void napi_schedule_irqoff(struct napi_struct *n) 434 { 435 if (napi_schedule_prep(n)) 436 __napi_schedule_irqoff(n); 437 } 438 439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 440 static inline bool napi_reschedule(struct napi_struct *napi) 441 { 442 if (napi_schedule_prep(napi)) { 443 __napi_schedule(napi); 444 return true; 445 } 446 return false; 447 } 448 449 void __napi_complete(struct napi_struct *n); 450 void napi_complete_done(struct napi_struct *n, int work_done); 451 /** 452 * napi_complete - NAPI processing complete 453 * @n: napi context 454 * 455 * Mark NAPI processing as complete. 456 * Consider using napi_complete_done() instead. 457 */ 458 static inline void napi_complete(struct napi_struct *n) 459 { 460 return napi_complete_done(n, 0); 461 } 462 463 /** 464 * napi_by_id - lookup a NAPI by napi_id 465 * @napi_id: hashed napi_id 466 * 467 * lookup @napi_id in napi_hash table 468 * must be called under rcu_read_lock() 469 */ 470 struct napi_struct *napi_by_id(unsigned int napi_id); 471 472 /** 473 * napi_hash_add - add a NAPI to global hashtable 474 * @napi: napi context 475 * 476 * generate a new napi_id and store a @napi under it in napi_hash 477 */ 478 void napi_hash_add(struct napi_struct *napi); 479 480 /** 481 * napi_hash_del - remove a NAPI from global table 482 * @napi: napi context 483 * 484 * Warning: caller must observe rcu grace period 485 * before freeing memory containing @napi 486 */ 487 void napi_hash_del(struct napi_struct *napi); 488 489 /** 490 * napi_disable - prevent NAPI from scheduling 491 * @n: napi context 492 * 493 * Stop NAPI from being scheduled on this context. 494 * Waits till any outstanding processing completes. 495 */ 496 void napi_disable(struct napi_struct *n); 497 498 /** 499 * napi_enable - enable NAPI scheduling 500 * @n: napi context 501 * 502 * Resume NAPI from being scheduled on this context. 503 * Must be paired with napi_disable. 504 */ 505 static inline void napi_enable(struct napi_struct *n) 506 { 507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 508 smp_mb__before_atomic(); 509 clear_bit(NAPI_STATE_SCHED, &n->state); 510 } 511 512 #ifdef CONFIG_SMP 513 /** 514 * napi_synchronize - wait until NAPI is not running 515 * @n: napi context 516 * 517 * Wait until NAPI is done being scheduled on this context. 518 * Waits till any outstanding processing completes but 519 * does not disable future activations. 520 */ 521 static inline void napi_synchronize(const struct napi_struct *n) 522 { 523 while (test_bit(NAPI_STATE_SCHED, &n->state)) 524 msleep(1); 525 } 526 #else 527 # define napi_synchronize(n) barrier() 528 #endif 529 530 enum netdev_queue_state_t { 531 __QUEUE_STATE_DRV_XOFF, 532 __QUEUE_STATE_STACK_XOFF, 533 __QUEUE_STATE_FROZEN, 534 }; 535 536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 539 540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 542 QUEUE_STATE_FROZEN) 543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 544 QUEUE_STATE_FROZEN) 545 546 /* 547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 548 * netif_tx_* functions below are used to manipulate this flag. The 549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 550 * queue independently. The netif_xmit_*stopped functions below are called 551 * to check if the queue has been stopped by the driver or stack (either 552 * of the XOFF bits are set in the state). Drivers should not need to call 553 * netif_xmit*stopped functions, they should only be using netif_tx_*. 554 */ 555 556 struct netdev_queue { 557 /* 558 * read mostly part 559 */ 560 struct net_device *dev; 561 struct Qdisc __rcu *qdisc; 562 struct Qdisc *qdisc_sleeping; 563 #ifdef CONFIG_SYSFS 564 struct kobject kobj; 565 #endif 566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 567 int numa_node; 568 #endif 569 /* 570 * write mostly part 571 */ 572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 573 int xmit_lock_owner; 574 /* 575 * please use this field instead of dev->trans_start 576 */ 577 unsigned long trans_start; 578 579 /* 580 * Number of TX timeouts for this queue 581 * (/sys/class/net/DEV/Q/trans_timeout) 582 */ 583 unsigned long trans_timeout; 584 585 unsigned long state; 586 587 #ifdef CONFIG_BQL 588 struct dql dql; 589 #endif 590 unsigned long tx_maxrate; 591 } ____cacheline_aligned_in_smp; 592 593 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 594 { 595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 596 return q->numa_node; 597 #else 598 return NUMA_NO_NODE; 599 #endif 600 } 601 602 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 603 { 604 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 605 q->numa_node = node; 606 #endif 607 } 608 609 #ifdef CONFIG_RPS 610 /* 611 * This structure holds an RPS map which can be of variable length. The 612 * map is an array of CPUs. 613 */ 614 struct rps_map { 615 unsigned int len; 616 struct rcu_head rcu; 617 u16 cpus[0]; 618 }; 619 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 620 621 /* 622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 623 * tail pointer for that CPU's input queue at the time of last enqueue, and 624 * a hardware filter index. 625 */ 626 struct rps_dev_flow { 627 u16 cpu; 628 u16 filter; 629 unsigned int last_qtail; 630 }; 631 #define RPS_NO_FILTER 0xffff 632 633 /* 634 * The rps_dev_flow_table structure contains a table of flow mappings. 635 */ 636 struct rps_dev_flow_table { 637 unsigned int mask; 638 struct rcu_head rcu; 639 struct rps_dev_flow flows[0]; 640 }; 641 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 642 ((_num) * sizeof(struct rps_dev_flow))) 643 644 /* 645 * The rps_sock_flow_table contains mappings of flows to the last CPU 646 * on which they were processed by the application (set in recvmsg). 647 * Each entry is a 32bit value. Upper part is the high order bits 648 * of flow hash, lower part is cpu number. 649 * rps_cpu_mask is used to partition the space, depending on number of 650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f, 652 * meaning we use 32-6=26 bits for the hash. 653 */ 654 struct rps_sock_flow_table { 655 u32 mask; 656 657 u32 ents[0] ____cacheline_aligned_in_smp; 658 }; 659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 660 661 #define RPS_NO_CPU 0xffff 662 663 extern u32 rps_cpu_mask; 664 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 665 666 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 667 u32 hash) 668 { 669 if (table && hash) { 670 unsigned int index = hash & table->mask; 671 u32 val = hash & ~rps_cpu_mask; 672 673 /* We only give a hint, preemption can change cpu under us */ 674 val |= raw_smp_processor_id(); 675 676 if (table->ents[index] != val) 677 table->ents[index] = val; 678 } 679 } 680 681 #ifdef CONFIG_RFS_ACCEL 682 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 683 u16 filter_id); 684 #endif 685 #endif /* CONFIG_RPS */ 686 687 /* This structure contains an instance of an RX queue. */ 688 struct netdev_rx_queue { 689 #ifdef CONFIG_RPS 690 struct rps_map __rcu *rps_map; 691 struct rps_dev_flow_table __rcu *rps_flow_table; 692 #endif 693 struct kobject kobj; 694 struct net_device *dev; 695 } ____cacheline_aligned_in_smp; 696 697 /* 698 * RX queue sysfs structures and functions. 699 */ 700 struct rx_queue_attribute { 701 struct attribute attr; 702 ssize_t (*show)(struct netdev_rx_queue *queue, 703 struct rx_queue_attribute *attr, char *buf); 704 ssize_t (*store)(struct netdev_rx_queue *queue, 705 struct rx_queue_attribute *attr, const char *buf, size_t len); 706 }; 707 708 #ifdef CONFIG_XPS 709 /* 710 * This structure holds an XPS map which can be of variable length. The 711 * map is an array of queues. 712 */ 713 struct xps_map { 714 unsigned int len; 715 unsigned int alloc_len; 716 struct rcu_head rcu; 717 u16 queues[0]; 718 }; 719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 721 / sizeof(u16)) 722 723 /* 724 * This structure holds all XPS maps for device. Maps are indexed by CPU. 725 */ 726 struct xps_dev_maps { 727 struct rcu_head rcu; 728 struct xps_map __rcu *cpu_map[0]; 729 }; 730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 731 (nr_cpu_ids * sizeof(struct xps_map *))) 732 #endif /* CONFIG_XPS */ 733 734 #define TC_MAX_QUEUE 16 735 #define TC_BITMASK 15 736 /* HW offloaded queuing disciplines txq count and offset maps */ 737 struct netdev_tc_txq { 738 u16 count; 739 u16 offset; 740 }; 741 742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 743 /* 744 * This structure is to hold information about the device 745 * configured to run FCoE protocol stack. 746 */ 747 struct netdev_fcoe_hbainfo { 748 char manufacturer[64]; 749 char serial_number[64]; 750 char hardware_version[64]; 751 char driver_version[64]; 752 char optionrom_version[64]; 753 char firmware_version[64]; 754 char model[256]; 755 char model_description[256]; 756 }; 757 #endif 758 759 #define MAX_PHYS_ITEM_ID_LEN 32 760 761 /* This structure holds a unique identifier to identify some 762 * physical item (port for example) used by a netdevice. 763 */ 764 struct netdev_phys_item_id { 765 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 766 unsigned char id_len; 767 }; 768 769 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 770 struct netdev_phys_item_id *b) 771 { 772 return a->id_len == b->id_len && 773 memcmp(a->id, b->id, a->id_len) == 0; 774 } 775 776 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 777 struct sk_buff *skb); 778 779 /* 780 * This structure defines the management hooks for network devices. 781 * The following hooks can be defined; unless noted otherwise, they are 782 * optional and can be filled with a null pointer. 783 * 784 * int (*ndo_init)(struct net_device *dev); 785 * This function is called once when network device is registered. 786 * The network device can use this to any late stage initializaton 787 * or semantic validattion. It can fail with an error code which will 788 * be propogated back to register_netdev 789 * 790 * void (*ndo_uninit)(struct net_device *dev); 791 * This function is called when device is unregistered or when registration 792 * fails. It is not called if init fails. 793 * 794 * int (*ndo_open)(struct net_device *dev); 795 * This function is called when network device transistions to the up 796 * state. 797 * 798 * int (*ndo_stop)(struct net_device *dev); 799 * This function is called when network device transistions to the down 800 * state. 801 * 802 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 803 * struct net_device *dev); 804 * Called when a packet needs to be transmitted. 805 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 806 * the queue before that can happen; it's for obsolete devices and weird 807 * corner cases, but the stack really does a non-trivial amount 808 * of useless work if you return NETDEV_TX_BUSY. 809 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 810 * Required can not be NULL. 811 * 812 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 813 * void *accel_priv, select_queue_fallback_t fallback); 814 * Called to decide which queue to when device supports multiple 815 * transmit queues. 816 * 817 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 818 * This function is called to allow device receiver to make 819 * changes to configuration when multicast or promiscious is enabled. 820 * 821 * void (*ndo_set_rx_mode)(struct net_device *dev); 822 * This function is called device changes address list filtering. 823 * If driver handles unicast address filtering, it should set 824 * IFF_UNICAST_FLT to its priv_flags. 825 * 826 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 827 * This function is called when the Media Access Control address 828 * needs to be changed. If this interface is not defined, the 829 * mac address can not be changed. 830 * 831 * int (*ndo_validate_addr)(struct net_device *dev); 832 * Test if Media Access Control address is valid for the device. 833 * 834 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 835 * Called when a user request an ioctl which can't be handled by 836 * the generic interface code. If not defined ioctl's return 837 * not supported error code. 838 * 839 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 840 * Used to set network devices bus interface parameters. This interface 841 * is retained for legacy reason, new devices should use the bus 842 * interface (PCI) for low level management. 843 * 844 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 845 * Called when a user wants to change the Maximum Transfer Unit 846 * of a device. If not defined, any request to change MTU will 847 * will return an error. 848 * 849 * void (*ndo_tx_timeout)(struct net_device *dev); 850 * Callback uses when the transmitter has not made any progress 851 * for dev->watchdog ticks. 852 * 853 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 854 * struct rtnl_link_stats64 *storage); 855 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 856 * Called when a user wants to get the network device usage 857 * statistics. Drivers must do one of the following: 858 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 859 * rtnl_link_stats64 structure passed by the caller. 860 * 2. Define @ndo_get_stats to update a net_device_stats structure 861 * (which should normally be dev->stats) and return a pointer to 862 * it. The structure may be changed asynchronously only if each 863 * field is written atomically. 864 * 3. Update dev->stats asynchronously and atomically, and define 865 * neither operation. 866 * 867 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 868 * If device support VLAN filtering this function is called when a 869 * VLAN id is registered. 870 * 871 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 872 * If device support VLAN filtering this function is called when a 873 * VLAN id is unregistered. 874 * 875 * void (*ndo_poll_controller)(struct net_device *dev); 876 * 877 * SR-IOV management functions. 878 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 879 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 880 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 881 * int max_tx_rate); 882 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 883 * int (*ndo_get_vf_config)(struct net_device *dev, 884 * int vf, struct ifla_vf_info *ivf); 885 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 886 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 887 * struct nlattr *port[]); 888 * 889 * Enable or disable the VF ability to query its RSS Redirection Table and 890 * Hash Key. This is needed since on some devices VF share this information 891 * with PF and querying it may adduce a theoretical security risk. 892 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 893 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 894 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 895 * Called to setup 'tc' number of traffic classes in the net device. This 896 * is always called from the stack with the rtnl lock held and netif tx 897 * queues stopped. This allows the netdevice to perform queue management 898 * safely. 899 * 900 * Fiber Channel over Ethernet (FCoE) offload functions. 901 * int (*ndo_fcoe_enable)(struct net_device *dev); 902 * Called when the FCoE protocol stack wants to start using LLD for FCoE 903 * so the underlying device can perform whatever needed configuration or 904 * initialization to support acceleration of FCoE traffic. 905 * 906 * int (*ndo_fcoe_disable)(struct net_device *dev); 907 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 908 * so the underlying device can perform whatever needed clean-ups to 909 * stop supporting acceleration of FCoE traffic. 910 * 911 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 912 * struct scatterlist *sgl, unsigned int sgc); 913 * Called when the FCoE Initiator wants to initialize an I/O that 914 * is a possible candidate for Direct Data Placement (DDP). The LLD can 915 * perform necessary setup and returns 1 to indicate the device is set up 916 * successfully to perform DDP on this I/O, otherwise this returns 0. 917 * 918 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 919 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 920 * indicated by the FC exchange id 'xid', so the underlying device can 921 * clean up and reuse resources for later DDP requests. 922 * 923 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 924 * struct scatterlist *sgl, unsigned int sgc); 925 * Called when the FCoE Target wants to initialize an I/O that 926 * is a possible candidate for Direct Data Placement (DDP). The LLD can 927 * perform necessary setup and returns 1 to indicate the device is set up 928 * successfully to perform DDP on this I/O, otherwise this returns 0. 929 * 930 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 931 * struct netdev_fcoe_hbainfo *hbainfo); 932 * Called when the FCoE Protocol stack wants information on the underlying 933 * device. This information is utilized by the FCoE protocol stack to 934 * register attributes with Fiber Channel management service as per the 935 * FC-GS Fabric Device Management Information(FDMI) specification. 936 * 937 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 938 * Called when the underlying device wants to override default World Wide 939 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 940 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 941 * protocol stack to use. 942 * 943 * RFS acceleration. 944 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 945 * u16 rxq_index, u32 flow_id); 946 * Set hardware filter for RFS. rxq_index is the target queue index; 947 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 948 * Return the filter ID on success, or a negative error code. 949 * 950 * Slave management functions (for bridge, bonding, etc). 951 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 952 * Called to make another netdev an underling. 953 * 954 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 955 * Called to release previously enslaved netdev. 956 * 957 * Feature/offload setting functions. 958 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 959 * netdev_features_t features); 960 * Adjusts the requested feature flags according to device-specific 961 * constraints, and returns the resulting flags. Must not modify 962 * the device state. 963 * 964 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 965 * Called to update device configuration to new features. Passed 966 * feature set might be less than what was returned by ndo_fix_features()). 967 * Must return >0 or -errno if it changed dev->features itself. 968 * 969 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 970 * struct net_device *dev, 971 * const unsigned char *addr, u16 vid, u16 flags) 972 * Adds an FDB entry to dev for addr. 973 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 974 * struct net_device *dev, 975 * const unsigned char *addr, u16 vid) 976 * Deletes the FDB entry from dev coresponding to addr. 977 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 978 * struct net_device *dev, struct net_device *filter_dev, 979 * int idx) 980 * Used to add FDB entries to dump requests. Implementers should add 981 * entries to skb and update idx with the number of entries. 982 * 983 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 984 * u16 flags) 985 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 986 * struct net_device *dev, u32 filter_mask, 987 * int nlflags) 988 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 989 * u16 flags); 990 * 991 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 992 * Called to change device carrier. Soft-devices (like dummy, team, etc) 993 * which do not represent real hardware may define this to allow their 994 * userspace components to manage their virtual carrier state. Devices 995 * that determine carrier state from physical hardware properties (eg 996 * network cables) or protocol-dependent mechanisms (eg 997 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 998 * 999 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1000 * struct netdev_phys_item_id *ppid); 1001 * Called to get ID of physical port of this device. If driver does 1002 * not implement this, it is assumed that the hw is not able to have 1003 * multiple net devices on single physical port. 1004 * 1005 * void (*ndo_add_vxlan_port)(struct net_device *dev, 1006 * sa_family_t sa_family, __be16 port); 1007 * Called by vxlan to notiy a driver about the UDP port and socket 1008 * address family that vxlan is listnening to. It is called only when 1009 * a new port starts listening. The operation is protected by the 1010 * vxlan_net->sock_lock. 1011 * 1012 * void (*ndo_del_vxlan_port)(struct net_device *dev, 1013 * sa_family_t sa_family, __be16 port); 1014 * Called by vxlan to notify the driver about a UDP port and socket 1015 * address family that vxlan is not listening to anymore. The operation 1016 * is protected by the vxlan_net->sock_lock. 1017 * 1018 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1019 * struct net_device *dev) 1020 * Called by upper layer devices to accelerate switching or other 1021 * station functionality into hardware. 'pdev is the lowerdev 1022 * to use for the offload and 'dev' is the net device that will 1023 * back the offload. Returns a pointer to the private structure 1024 * the upper layer will maintain. 1025 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1026 * Called by upper layer device to delete the station created 1027 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1028 * the station and priv is the structure returned by the add 1029 * operation. 1030 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, 1031 * struct net_device *dev, 1032 * void *priv); 1033 * Callback to use for xmit over the accelerated station. This 1034 * is used in place of ndo_start_xmit on accelerated net 1035 * devices. 1036 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1037 * struct net_device *dev 1038 * netdev_features_t features); 1039 * Called by core transmit path to determine if device is capable of 1040 * performing offload operations on a given packet. This is to give 1041 * the device an opportunity to implement any restrictions that cannot 1042 * be otherwise expressed by feature flags. The check is called with 1043 * the set of features that the stack has calculated and it returns 1044 * those the driver believes to be appropriate. 1045 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1046 * int queue_index, u32 maxrate); 1047 * Called when a user wants to set a max-rate limitation of specific 1048 * TX queue. 1049 * int (*ndo_get_iflink)(const struct net_device *dev); 1050 * Called to get the iflink value of this device. 1051 * void (*ndo_change_proto_down)(struct net_device *dev, 1052 * bool proto_down); 1053 * This function is used to pass protocol port error state information 1054 * to the switch driver. The switch driver can react to the proto_down 1055 * by doing a phys down on the associated switch port. 1056 * 1057 */ 1058 struct net_device_ops { 1059 int (*ndo_init)(struct net_device *dev); 1060 void (*ndo_uninit)(struct net_device *dev); 1061 int (*ndo_open)(struct net_device *dev); 1062 int (*ndo_stop)(struct net_device *dev); 1063 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 1064 struct net_device *dev); 1065 u16 (*ndo_select_queue)(struct net_device *dev, 1066 struct sk_buff *skb, 1067 void *accel_priv, 1068 select_queue_fallback_t fallback); 1069 void (*ndo_change_rx_flags)(struct net_device *dev, 1070 int flags); 1071 void (*ndo_set_rx_mode)(struct net_device *dev); 1072 int (*ndo_set_mac_address)(struct net_device *dev, 1073 void *addr); 1074 int (*ndo_validate_addr)(struct net_device *dev); 1075 int (*ndo_do_ioctl)(struct net_device *dev, 1076 struct ifreq *ifr, int cmd); 1077 int (*ndo_set_config)(struct net_device *dev, 1078 struct ifmap *map); 1079 int (*ndo_change_mtu)(struct net_device *dev, 1080 int new_mtu); 1081 int (*ndo_neigh_setup)(struct net_device *dev, 1082 struct neigh_parms *); 1083 void (*ndo_tx_timeout) (struct net_device *dev); 1084 1085 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 1086 struct rtnl_link_stats64 *storage); 1087 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1088 1089 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1090 __be16 proto, u16 vid); 1091 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1092 __be16 proto, u16 vid); 1093 #ifdef CONFIG_NET_POLL_CONTROLLER 1094 void (*ndo_poll_controller)(struct net_device *dev); 1095 int (*ndo_netpoll_setup)(struct net_device *dev, 1096 struct netpoll_info *info); 1097 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1098 #endif 1099 #ifdef CONFIG_NET_RX_BUSY_POLL 1100 int (*ndo_busy_poll)(struct napi_struct *dev); 1101 #endif 1102 int (*ndo_set_vf_mac)(struct net_device *dev, 1103 int queue, u8 *mac); 1104 int (*ndo_set_vf_vlan)(struct net_device *dev, 1105 int queue, u16 vlan, u8 qos); 1106 int (*ndo_set_vf_rate)(struct net_device *dev, 1107 int vf, int min_tx_rate, 1108 int max_tx_rate); 1109 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1110 int vf, bool setting); 1111 int (*ndo_get_vf_config)(struct net_device *dev, 1112 int vf, 1113 struct ifla_vf_info *ivf); 1114 int (*ndo_set_vf_link_state)(struct net_device *dev, 1115 int vf, int link_state); 1116 int (*ndo_get_vf_stats)(struct net_device *dev, 1117 int vf, 1118 struct ifla_vf_stats 1119 *vf_stats); 1120 int (*ndo_set_vf_port)(struct net_device *dev, 1121 int vf, 1122 struct nlattr *port[]); 1123 int (*ndo_get_vf_port)(struct net_device *dev, 1124 int vf, struct sk_buff *skb); 1125 int (*ndo_set_vf_rss_query_en)( 1126 struct net_device *dev, 1127 int vf, bool setting); 1128 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 1129 #if IS_ENABLED(CONFIG_FCOE) 1130 int (*ndo_fcoe_enable)(struct net_device *dev); 1131 int (*ndo_fcoe_disable)(struct net_device *dev); 1132 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1133 u16 xid, 1134 struct scatterlist *sgl, 1135 unsigned int sgc); 1136 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1137 u16 xid); 1138 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1139 u16 xid, 1140 struct scatterlist *sgl, 1141 unsigned int sgc); 1142 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1143 struct netdev_fcoe_hbainfo *hbainfo); 1144 #endif 1145 1146 #if IS_ENABLED(CONFIG_LIBFCOE) 1147 #define NETDEV_FCOE_WWNN 0 1148 #define NETDEV_FCOE_WWPN 1 1149 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1150 u64 *wwn, int type); 1151 #endif 1152 1153 #ifdef CONFIG_RFS_ACCEL 1154 int (*ndo_rx_flow_steer)(struct net_device *dev, 1155 const struct sk_buff *skb, 1156 u16 rxq_index, 1157 u32 flow_id); 1158 #endif 1159 int (*ndo_add_slave)(struct net_device *dev, 1160 struct net_device *slave_dev); 1161 int (*ndo_del_slave)(struct net_device *dev, 1162 struct net_device *slave_dev); 1163 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1164 netdev_features_t features); 1165 int (*ndo_set_features)(struct net_device *dev, 1166 netdev_features_t features); 1167 int (*ndo_neigh_construct)(struct neighbour *n); 1168 void (*ndo_neigh_destroy)(struct neighbour *n); 1169 1170 int (*ndo_fdb_add)(struct ndmsg *ndm, 1171 struct nlattr *tb[], 1172 struct net_device *dev, 1173 const unsigned char *addr, 1174 u16 vid, 1175 u16 flags); 1176 int (*ndo_fdb_del)(struct ndmsg *ndm, 1177 struct nlattr *tb[], 1178 struct net_device *dev, 1179 const unsigned char *addr, 1180 u16 vid); 1181 int (*ndo_fdb_dump)(struct sk_buff *skb, 1182 struct netlink_callback *cb, 1183 struct net_device *dev, 1184 struct net_device *filter_dev, 1185 int idx); 1186 1187 int (*ndo_bridge_setlink)(struct net_device *dev, 1188 struct nlmsghdr *nlh, 1189 u16 flags); 1190 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1191 u32 pid, u32 seq, 1192 struct net_device *dev, 1193 u32 filter_mask, 1194 int nlflags); 1195 int (*ndo_bridge_dellink)(struct net_device *dev, 1196 struct nlmsghdr *nlh, 1197 u16 flags); 1198 int (*ndo_change_carrier)(struct net_device *dev, 1199 bool new_carrier); 1200 int (*ndo_get_phys_port_id)(struct net_device *dev, 1201 struct netdev_phys_item_id *ppid); 1202 int (*ndo_get_phys_port_name)(struct net_device *dev, 1203 char *name, size_t len); 1204 void (*ndo_add_vxlan_port)(struct net_device *dev, 1205 sa_family_t sa_family, 1206 __be16 port); 1207 void (*ndo_del_vxlan_port)(struct net_device *dev, 1208 sa_family_t sa_family, 1209 __be16 port); 1210 1211 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1212 struct net_device *dev); 1213 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1214 void *priv); 1215 1216 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, 1217 struct net_device *dev, 1218 void *priv); 1219 int (*ndo_get_lock_subclass)(struct net_device *dev); 1220 netdev_features_t (*ndo_features_check) (struct sk_buff *skb, 1221 struct net_device *dev, 1222 netdev_features_t features); 1223 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1224 int queue_index, 1225 u32 maxrate); 1226 int (*ndo_get_iflink)(const struct net_device *dev); 1227 int (*ndo_change_proto_down)(struct net_device *dev, 1228 bool proto_down); 1229 }; 1230 1231 /** 1232 * enum net_device_priv_flags - &struct net_device priv_flags 1233 * 1234 * These are the &struct net_device, they are only set internally 1235 * by drivers and used in the kernel. These flags are invisible to 1236 * userspace, this means that the order of these flags can change 1237 * during any kernel release. 1238 * 1239 * You should have a pretty good reason to be extending these flags. 1240 * 1241 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1242 * @IFF_EBRIDGE: Ethernet bridging device 1243 * @IFF_BONDING: bonding master or slave 1244 * @IFF_ISATAP: ISATAP interface (RFC4214) 1245 * @IFF_WAN_HDLC: WAN HDLC device 1246 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1247 * release skb->dst 1248 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1249 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1250 * @IFF_MACVLAN_PORT: device used as macvlan port 1251 * @IFF_BRIDGE_PORT: device used as bridge port 1252 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1253 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1254 * @IFF_UNICAST_FLT: Supports unicast filtering 1255 * @IFF_TEAM_PORT: device used as team port 1256 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1257 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1258 * change when it's running 1259 * @IFF_MACVLAN: Macvlan device 1260 * @IFF_VRF_MASTER: device is a VRF master 1261 * @IFF_NO_QUEUE: device can run without qdisc attached 1262 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1263 */ 1264 enum netdev_priv_flags { 1265 IFF_802_1Q_VLAN = 1<<0, 1266 IFF_EBRIDGE = 1<<1, 1267 IFF_BONDING = 1<<2, 1268 IFF_ISATAP = 1<<3, 1269 IFF_WAN_HDLC = 1<<4, 1270 IFF_XMIT_DST_RELEASE = 1<<5, 1271 IFF_DONT_BRIDGE = 1<<6, 1272 IFF_DISABLE_NETPOLL = 1<<7, 1273 IFF_MACVLAN_PORT = 1<<8, 1274 IFF_BRIDGE_PORT = 1<<9, 1275 IFF_OVS_DATAPATH = 1<<10, 1276 IFF_TX_SKB_SHARING = 1<<11, 1277 IFF_UNICAST_FLT = 1<<12, 1278 IFF_TEAM_PORT = 1<<13, 1279 IFF_SUPP_NOFCS = 1<<14, 1280 IFF_LIVE_ADDR_CHANGE = 1<<15, 1281 IFF_MACVLAN = 1<<16, 1282 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1283 IFF_IPVLAN_MASTER = 1<<18, 1284 IFF_IPVLAN_SLAVE = 1<<19, 1285 IFF_VRF_MASTER = 1<<20, 1286 IFF_NO_QUEUE = 1<<21, 1287 IFF_OPENVSWITCH = 1<<22, 1288 }; 1289 1290 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1291 #define IFF_EBRIDGE IFF_EBRIDGE 1292 #define IFF_BONDING IFF_BONDING 1293 #define IFF_ISATAP IFF_ISATAP 1294 #define IFF_WAN_HDLC IFF_WAN_HDLC 1295 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1296 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1297 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1298 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1299 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1300 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1301 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1302 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1303 #define IFF_TEAM_PORT IFF_TEAM_PORT 1304 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1305 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1306 #define IFF_MACVLAN IFF_MACVLAN 1307 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1308 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1309 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1310 #define IFF_VRF_MASTER IFF_VRF_MASTER 1311 #define IFF_NO_QUEUE IFF_NO_QUEUE 1312 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1313 1314 /** 1315 * struct net_device - The DEVICE structure. 1316 * Actually, this whole structure is a big mistake. It mixes I/O 1317 * data with strictly "high-level" data, and it has to know about 1318 * almost every data structure used in the INET module. 1319 * 1320 * @name: This is the first field of the "visible" part of this structure 1321 * (i.e. as seen by users in the "Space.c" file). It is the name 1322 * of the interface. 1323 * 1324 * @name_hlist: Device name hash chain, please keep it close to name[] 1325 * @ifalias: SNMP alias 1326 * @mem_end: Shared memory end 1327 * @mem_start: Shared memory start 1328 * @base_addr: Device I/O address 1329 * @irq: Device IRQ number 1330 * 1331 * @carrier_changes: Stats to monitor carrier on<->off transitions 1332 * 1333 * @state: Generic network queuing layer state, see netdev_state_t 1334 * @dev_list: The global list of network devices 1335 * @napi_list: List entry, that is used for polling napi devices 1336 * @unreg_list: List entry, that is used, when we are unregistering the 1337 * device, see the function unregister_netdev 1338 * @close_list: List entry, that is used, when we are closing the device 1339 * 1340 * @adj_list: Directly linked devices, like slaves for bonding 1341 * @all_adj_list: All linked devices, *including* neighbours 1342 * @features: Currently active device features 1343 * @hw_features: User-changeable features 1344 * 1345 * @wanted_features: User-requested features 1346 * @vlan_features: Mask of features inheritable by VLAN devices 1347 * 1348 * @hw_enc_features: Mask of features inherited by encapsulating devices 1349 * This field indicates what encapsulation 1350 * offloads the hardware is capable of doing, 1351 * and drivers will need to set them appropriately. 1352 * 1353 * @mpls_features: Mask of features inheritable by MPLS 1354 * 1355 * @ifindex: interface index 1356 * @group: The group, that the device belongs to 1357 * 1358 * @stats: Statistics struct, which was left as a legacy, use 1359 * rtnl_link_stats64 instead 1360 * 1361 * @rx_dropped: Dropped packets by core network, 1362 * do not use this in drivers 1363 * @tx_dropped: Dropped packets by core network, 1364 * do not use this in drivers 1365 * 1366 * @wireless_handlers: List of functions to handle Wireless Extensions, 1367 * instead of ioctl, 1368 * see <net/iw_handler.h> for details. 1369 * @wireless_data: Instance data managed by the core of wireless extensions 1370 * 1371 * @netdev_ops: Includes several pointers to callbacks, 1372 * if one wants to override the ndo_*() functions 1373 * @ethtool_ops: Management operations 1374 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1375 * of Layer 2 headers. 1376 * 1377 * @flags: Interface flags (a la BSD) 1378 * @priv_flags: Like 'flags' but invisible to userspace, 1379 * see if.h for the definitions 1380 * @gflags: Global flags ( kept as legacy ) 1381 * @padded: How much padding added by alloc_netdev() 1382 * @operstate: RFC2863 operstate 1383 * @link_mode: Mapping policy to operstate 1384 * @if_port: Selectable AUI, TP, ... 1385 * @dma: DMA channel 1386 * @mtu: Interface MTU value 1387 * @type: Interface hardware type 1388 * @hard_header_len: Hardware header length 1389 * 1390 * @needed_headroom: Extra headroom the hardware may need, but not in all 1391 * cases can this be guaranteed 1392 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1393 * cases can this be guaranteed. Some cases also use 1394 * LL_MAX_HEADER instead to allocate the skb 1395 * 1396 * interface address info: 1397 * 1398 * @perm_addr: Permanent hw address 1399 * @addr_assign_type: Hw address assignment type 1400 * @addr_len: Hardware address length 1401 * @neigh_priv_len; Used in neigh_alloc(), 1402 * initialized only in atm/clip.c 1403 * @dev_id: Used to differentiate devices that share 1404 * the same link layer address 1405 * @dev_port: Used to differentiate devices that share 1406 * the same function 1407 * @addr_list_lock: XXX: need comments on this one 1408 * @uc_promisc: Counter, that indicates, that promiscuous mode 1409 * has been enabled due to the need to listen to 1410 * additional unicast addresses in a device that 1411 * does not implement ndo_set_rx_mode() 1412 * @uc: unicast mac addresses 1413 * @mc: multicast mac addresses 1414 * @dev_addrs: list of device hw addresses 1415 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1416 * @promiscuity: Number of times, the NIC is told to work in 1417 * Promiscuous mode, if it becomes 0 the NIC will 1418 * exit from working in Promiscuous mode 1419 * @allmulti: Counter, enables or disables allmulticast mode 1420 * 1421 * @vlan_info: VLAN info 1422 * @dsa_ptr: dsa specific data 1423 * @tipc_ptr: TIPC specific data 1424 * @atalk_ptr: AppleTalk link 1425 * @ip_ptr: IPv4 specific data 1426 * @dn_ptr: DECnet specific data 1427 * @ip6_ptr: IPv6 specific data 1428 * @ax25_ptr: AX.25 specific data 1429 * @vrf_ptr: VRF specific data 1430 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1431 * 1432 * @last_rx: Time of last Rx 1433 * @dev_addr: Hw address (before bcast, 1434 * because most packets are unicast) 1435 * 1436 * @_rx: Array of RX queues 1437 * @num_rx_queues: Number of RX queues 1438 * allocated at register_netdev() time 1439 * @real_num_rx_queues: Number of RX queues currently active in device 1440 * 1441 * @rx_handler: handler for received packets 1442 * @rx_handler_data: XXX: need comments on this one 1443 * @ingress_queue: XXX: need comments on this one 1444 * @broadcast: hw bcast address 1445 * 1446 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1447 * indexed by RX queue number. Assigned by driver. 1448 * This must only be set if the ndo_rx_flow_steer 1449 * operation is defined 1450 * @index_hlist: Device index hash chain 1451 * 1452 * @_tx: Array of TX queues 1453 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1454 * @real_num_tx_queues: Number of TX queues currently active in device 1455 * @qdisc: Root qdisc from userspace point of view 1456 * @tx_queue_len: Max frames per queue allowed 1457 * @tx_global_lock: XXX: need comments on this one 1458 * 1459 * @xps_maps: XXX: need comments on this one 1460 * 1461 * @offload_fwd_mark: Offload device fwding mark 1462 * 1463 * @trans_start: Time (in jiffies) of last Tx 1464 * @watchdog_timeo: Represents the timeout that is used by 1465 * the watchdog ( see dev_watchdog() ) 1466 * @watchdog_timer: List of timers 1467 * 1468 * @pcpu_refcnt: Number of references to this device 1469 * @todo_list: Delayed register/unregister 1470 * @link_watch_list: XXX: need comments on this one 1471 * 1472 * @reg_state: Register/unregister state machine 1473 * @dismantle: Device is going to be freed 1474 * @rtnl_link_state: This enum represents the phases of creating 1475 * a new link 1476 * 1477 * @destructor: Called from unregister, 1478 * can be used to call free_netdev 1479 * @npinfo: XXX: need comments on this one 1480 * @nd_net: Network namespace this network device is inside 1481 * 1482 * @ml_priv: Mid-layer private 1483 * @lstats: Loopback statistics 1484 * @tstats: Tunnel statistics 1485 * @dstats: Dummy statistics 1486 * @vstats: Virtual ethernet statistics 1487 * 1488 * @garp_port: GARP 1489 * @mrp_port: MRP 1490 * 1491 * @dev: Class/net/name entry 1492 * @sysfs_groups: Space for optional device, statistics and wireless 1493 * sysfs groups 1494 * 1495 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1496 * @rtnl_link_ops: Rtnl_link_ops 1497 * 1498 * @gso_max_size: Maximum size of generic segmentation offload 1499 * @gso_max_segs: Maximum number of segments that can be passed to the 1500 * NIC for GSO 1501 * @gso_min_segs: Minimum number of segments that can be passed to the 1502 * NIC for GSO 1503 * 1504 * @dcbnl_ops: Data Center Bridging netlink ops 1505 * @num_tc: Number of traffic classes in the net device 1506 * @tc_to_txq: XXX: need comments on this one 1507 * @prio_tc_map XXX: need comments on this one 1508 * 1509 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1510 * 1511 * @priomap: XXX: need comments on this one 1512 * @phydev: Physical device may attach itself 1513 * for hardware timestamping 1514 * 1515 * @qdisc_tx_busylock: XXX: need comments on this one 1516 * 1517 * @proto_down: protocol port state information can be sent to the 1518 * switch driver and used to set the phys state of the 1519 * switch port. 1520 * 1521 * FIXME: cleanup struct net_device such that network protocol info 1522 * moves out. 1523 */ 1524 1525 struct net_device { 1526 char name[IFNAMSIZ]; 1527 struct hlist_node name_hlist; 1528 char *ifalias; 1529 /* 1530 * I/O specific fields 1531 * FIXME: Merge these and struct ifmap into one 1532 */ 1533 unsigned long mem_end; 1534 unsigned long mem_start; 1535 unsigned long base_addr; 1536 int irq; 1537 1538 atomic_t carrier_changes; 1539 1540 /* 1541 * Some hardware also needs these fields (state,dev_list, 1542 * napi_list,unreg_list,close_list) but they are not 1543 * part of the usual set specified in Space.c. 1544 */ 1545 1546 unsigned long state; 1547 1548 struct list_head dev_list; 1549 struct list_head napi_list; 1550 struct list_head unreg_list; 1551 struct list_head close_list; 1552 struct list_head ptype_all; 1553 struct list_head ptype_specific; 1554 1555 struct { 1556 struct list_head upper; 1557 struct list_head lower; 1558 } adj_list; 1559 1560 struct { 1561 struct list_head upper; 1562 struct list_head lower; 1563 } all_adj_list; 1564 1565 netdev_features_t features; 1566 netdev_features_t hw_features; 1567 netdev_features_t wanted_features; 1568 netdev_features_t vlan_features; 1569 netdev_features_t hw_enc_features; 1570 netdev_features_t mpls_features; 1571 1572 int ifindex; 1573 int group; 1574 1575 struct net_device_stats stats; 1576 1577 atomic_long_t rx_dropped; 1578 atomic_long_t tx_dropped; 1579 1580 #ifdef CONFIG_WIRELESS_EXT 1581 const struct iw_handler_def * wireless_handlers; 1582 struct iw_public_data * wireless_data; 1583 #endif 1584 const struct net_device_ops *netdev_ops; 1585 const struct ethtool_ops *ethtool_ops; 1586 #ifdef CONFIG_NET_SWITCHDEV 1587 const struct switchdev_ops *switchdev_ops; 1588 #endif 1589 1590 const struct header_ops *header_ops; 1591 1592 unsigned int flags; 1593 unsigned int priv_flags; 1594 1595 unsigned short gflags; 1596 unsigned short padded; 1597 1598 unsigned char operstate; 1599 unsigned char link_mode; 1600 1601 unsigned char if_port; 1602 unsigned char dma; 1603 1604 unsigned int mtu; 1605 unsigned short type; 1606 unsigned short hard_header_len; 1607 1608 unsigned short needed_headroom; 1609 unsigned short needed_tailroom; 1610 1611 /* Interface address info. */ 1612 unsigned char perm_addr[MAX_ADDR_LEN]; 1613 unsigned char addr_assign_type; 1614 unsigned char addr_len; 1615 unsigned short neigh_priv_len; 1616 unsigned short dev_id; 1617 unsigned short dev_port; 1618 spinlock_t addr_list_lock; 1619 unsigned char name_assign_type; 1620 bool uc_promisc; 1621 struct netdev_hw_addr_list uc; 1622 struct netdev_hw_addr_list mc; 1623 struct netdev_hw_addr_list dev_addrs; 1624 1625 #ifdef CONFIG_SYSFS 1626 struct kset *queues_kset; 1627 #endif 1628 unsigned int promiscuity; 1629 unsigned int allmulti; 1630 1631 1632 /* Protocol specific pointers */ 1633 1634 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1635 struct vlan_info __rcu *vlan_info; 1636 #endif 1637 #if IS_ENABLED(CONFIG_NET_DSA) 1638 struct dsa_switch_tree *dsa_ptr; 1639 #endif 1640 #if IS_ENABLED(CONFIG_TIPC) 1641 struct tipc_bearer __rcu *tipc_ptr; 1642 #endif 1643 void *atalk_ptr; 1644 struct in_device __rcu *ip_ptr; 1645 struct dn_dev __rcu *dn_ptr; 1646 struct inet6_dev __rcu *ip6_ptr; 1647 void *ax25_ptr; 1648 struct net_vrf_dev __rcu *vrf_ptr; 1649 struct wireless_dev *ieee80211_ptr; 1650 struct wpan_dev *ieee802154_ptr; 1651 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1652 struct mpls_dev __rcu *mpls_ptr; 1653 #endif 1654 1655 /* 1656 * Cache lines mostly used on receive path (including eth_type_trans()) 1657 */ 1658 unsigned long last_rx; 1659 1660 /* Interface address info used in eth_type_trans() */ 1661 unsigned char *dev_addr; 1662 1663 1664 #ifdef CONFIG_SYSFS 1665 struct netdev_rx_queue *_rx; 1666 1667 unsigned int num_rx_queues; 1668 unsigned int real_num_rx_queues; 1669 1670 #endif 1671 1672 unsigned long gro_flush_timeout; 1673 rx_handler_func_t __rcu *rx_handler; 1674 void __rcu *rx_handler_data; 1675 1676 #ifdef CONFIG_NET_CLS_ACT 1677 struct tcf_proto __rcu *ingress_cl_list; 1678 #endif 1679 struct netdev_queue __rcu *ingress_queue; 1680 #ifdef CONFIG_NETFILTER_INGRESS 1681 struct list_head nf_hooks_ingress; 1682 #endif 1683 1684 unsigned char broadcast[MAX_ADDR_LEN]; 1685 #ifdef CONFIG_RFS_ACCEL 1686 struct cpu_rmap *rx_cpu_rmap; 1687 #endif 1688 struct hlist_node index_hlist; 1689 1690 /* 1691 * Cache lines mostly used on transmit path 1692 */ 1693 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1694 unsigned int num_tx_queues; 1695 unsigned int real_num_tx_queues; 1696 struct Qdisc *qdisc; 1697 unsigned long tx_queue_len; 1698 spinlock_t tx_global_lock; 1699 int watchdog_timeo; 1700 1701 #ifdef CONFIG_XPS 1702 struct xps_dev_maps __rcu *xps_maps; 1703 #endif 1704 1705 #ifdef CONFIG_NET_SWITCHDEV 1706 u32 offload_fwd_mark; 1707 #endif 1708 1709 /* These may be needed for future network-power-down code. */ 1710 1711 /* 1712 * trans_start here is expensive for high speed devices on SMP, 1713 * please use netdev_queue->trans_start instead. 1714 */ 1715 unsigned long trans_start; 1716 1717 struct timer_list watchdog_timer; 1718 1719 int __percpu *pcpu_refcnt; 1720 struct list_head todo_list; 1721 1722 struct list_head link_watch_list; 1723 1724 enum { NETREG_UNINITIALIZED=0, 1725 NETREG_REGISTERED, /* completed register_netdevice */ 1726 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1727 NETREG_UNREGISTERED, /* completed unregister todo */ 1728 NETREG_RELEASED, /* called free_netdev */ 1729 NETREG_DUMMY, /* dummy device for NAPI poll */ 1730 } reg_state:8; 1731 1732 bool dismantle; 1733 1734 enum { 1735 RTNL_LINK_INITIALIZED, 1736 RTNL_LINK_INITIALIZING, 1737 } rtnl_link_state:16; 1738 1739 void (*destructor)(struct net_device *dev); 1740 1741 #ifdef CONFIG_NETPOLL 1742 struct netpoll_info __rcu *npinfo; 1743 #endif 1744 1745 possible_net_t nd_net; 1746 1747 /* mid-layer private */ 1748 union { 1749 void *ml_priv; 1750 struct pcpu_lstats __percpu *lstats; 1751 struct pcpu_sw_netstats __percpu *tstats; 1752 struct pcpu_dstats __percpu *dstats; 1753 struct pcpu_vstats __percpu *vstats; 1754 }; 1755 1756 struct garp_port __rcu *garp_port; 1757 struct mrp_port __rcu *mrp_port; 1758 1759 struct device dev; 1760 const struct attribute_group *sysfs_groups[4]; 1761 const struct attribute_group *sysfs_rx_queue_group; 1762 1763 const struct rtnl_link_ops *rtnl_link_ops; 1764 1765 /* for setting kernel sock attribute on TCP connection setup */ 1766 #define GSO_MAX_SIZE 65536 1767 unsigned int gso_max_size; 1768 #define GSO_MAX_SEGS 65535 1769 u16 gso_max_segs; 1770 u16 gso_min_segs; 1771 #ifdef CONFIG_DCB 1772 const struct dcbnl_rtnl_ops *dcbnl_ops; 1773 #endif 1774 u8 num_tc; 1775 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1776 u8 prio_tc_map[TC_BITMASK + 1]; 1777 1778 #if IS_ENABLED(CONFIG_FCOE) 1779 unsigned int fcoe_ddp_xid; 1780 #endif 1781 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1782 struct netprio_map __rcu *priomap; 1783 #endif 1784 struct phy_device *phydev; 1785 struct lock_class_key *qdisc_tx_busylock; 1786 bool proto_down; 1787 }; 1788 #define to_net_dev(d) container_of(d, struct net_device, dev) 1789 1790 #define NETDEV_ALIGN 32 1791 1792 static inline 1793 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1794 { 1795 return dev->prio_tc_map[prio & TC_BITMASK]; 1796 } 1797 1798 static inline 1799 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1800 { 1801 if (tc >= dev->num_tc) 1802 return -EINVAL; 1803 1804 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1805 return 0; 1806 } 1807 1808 static inline 1809 void netdev_reset_tc(struct net_device *dev) 1810 { 1811 dev->num_tc = 0; 1812 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1813 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1814 } 1815 1816 static inline 1817 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1818 { 1819 if (tc >= dev->num_tc) 1820 return -EINVAL; 1821 1822 dev->tc_to_txq[tc].count = count; 1823 dev->tc_to_txq[tc].offset = offset; 1824 return 0; 1825 } 1826 1827 static inline 1828 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1829 { 1830 if (num_tc > TC_MAX_QUEUE) 1831 return -EINVAL; 1832 1833 dev->num_tc = num_tc; 1834 return 0; 1835 } 1836 1837 static inline 1838 int netdev_get_num_tc(struct net_device *dev) 1839 { 1840 return dev->num_tc; 1841 } 1842 1843 static inline 1844 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1845 unsigned int index) 1846 { 1847 return &dev->_tx[index]; 1848 } 1849 1850 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1851 const struct sk_buff *skb) 1852 { 1853 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1854 } 1855 1856 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1857 void (*f)(struct net_device *, 1858 struct netdev_queue *, 1859 void *), 1860 void *arg) 1861 { 1862 unsigned int i; 1863 1864 for (i = 0; i < dev->num_tx_queues; i++) 1865 f(dev, &dev->_tx[i], arg); 1866 } 1867 1868 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1869 struct sk_buff *skb, 1870 void *accel_priv); 1871 1872 /* 1873 * Net namespace inlines 1874 */ 1875 static inline 1876 struct net *dev_net(const struct net_device *dev) 1877 { 1878 return read_pnet(&dev->nd_net); 1879 } 1880 1881 static inline 1882 void dev_net_set(struct net_device *dev, struct net *net) 1883 { 1884 write_pnet(&dev->nd_net, net); 1885 } 1886 1887 static inline bool netdev_uses_dsa(struct net_device *dev) 1888 { 1889 #if IS_ENABLED(CONFIG_NET_DSA) 1890 if (dev->dsa_ptr != NULL) 1891 return dsa_uses_tagged_protocol(dev->dsa_ptr); 1892 #endif 1893 return false; 1894 } 1895 1896 /** 1897 * netdev_priv - access network device private data 1898 * @dev: network device 1899 * 1900 * Get network device private data 1901 */ 1902 static inline void *netdev_priv(const struct net_device *dev) 1903 { 1904 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1905 } 1906 1907 /* Set the sysfs physical device reference for the network logical device 1908 * if set prior to registration will cause a symlink during initialization. 1909 */ 1910 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1911 1912 /* Set the sysfs device type for the network logical device to allow 1913 * fine-grained identification of different network device types. For 1914 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1915 */ 1916 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1917 1918 /* Default NAPI poll() weight 1919 * Device drivers are strongly advised to not use bigger value 1920 */ 1921 #define NAPI_POLL_WEIGHT 64 1922 1923 /** 1924 * netif_napi_add - initialize a napi context 1925 * @dev: network device 1926 * @napi: napi context 1927 * @poll: polling function 1928 * @weight: default weight 1929 * 1930 * netif_napi_add() must be used to initialize a napi context prior to calling 1931 * *any* of the other napi related functions. 1932 */ 1933 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1934 int (*poll)(struct napi_struct *, int), int weight); 1935 1936 /** 1937 * netif_napi_del - remove a napi context 1938 * @napi: napi context 1939 * 1940 * netif_napi_del() removes a napi context from the network device napi list 1941 */ 1942 void netif_napi_del(struct napi_struct *napi); 1943 1944 struct napi_gro_cb { 1945 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1946 void *frag0; 1947 1948 /* Length of frag0. */ 1949 unsigned int frag0_len; 1950 1951 /* This indicates where we are processing relative to skb->data. */ 1952 int data_offset; 1953 1954 /* This is non-zero if the packet cannot be merged with the new skb. */ 1955 u16 flush; 1956 1957 /* Save the IP ID here and check when we get to the transport layer */ 1958 u16 flush_id; 1959 1960 /* Number of segments aggregated. */ 1961 u16 count; 1962 1963 /* Start offset for remote checksum offload */ 1964 u16 gro_remcsum_start; 1965 1966 /* jiffies when first packet was created/queued */ 1967 unsigned long age; 1968 1969 /* Used in ipv6_gro_receive() and foo-over-udp */ 1970 u16 proto; 1971 1972 /* This is non-zero if the packet may be of the same flow. */ 1973 u8 same_flow:1; 1974 1975 /* Used in udp_gro_receive */ 1976 u8 udp_mark:1; 1977 1978 /* GRO checksum is valid */ 1979 u8 csum_valid:1; 1980 1981 /* Number of checksums via CHECKSUM_UNNECESSARY */ 1982 u8 csum_cnt:3; 1983 1984 /* Free the skb? */ 1985 u8 free:2; 1986 #define NAPI_GRO_FREE 1 1987 #define NAPI_GRO_FREE_STOLEN_HEAD 2 1988 1989 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 1990 u8 is_ipv6:1; 1991 1992 /* 7 bit hole */ 1993 1994 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 1995 __wsum csum; 1996 1997 /* used in skb_gro_receive() slow path */ 1998 struct sk_buff *last; 1999 }; 2000 2001 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2002 2003 struct packet_type { 2004 __be16 type; /* This is really htons(ether_type). */ 2005 struct net_device *dev; /* NULL is wildcarded here */ 2006 int (*func) (struct sk_buff *, 2007 struct net_device *, 2008 struct packet_type *, 2009 struct net_device *); 2010 bool (*id_match)(struct packet_type *ptype, 2011 struct sock *sk); 2012 void *af_packet_priv; 2013 struct list_head list; 2014 }; 2015 2016 struct offload_callbacks { 2017 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2018 netdev_features_t features); 2019 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2020 struct sk_buff *skb); 2021 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2022 }; 2023 2024 struct packet_offload { 2025 __be16 type; /* This is really htons(ether_type). */ 2026 u16 priority; 2027 struct offload_callbacks callbacks; 2028 struct list_head list; 2029 }; 2030 2031 struct udp_offload; 2032 2033 struct udp_offload_callbacks { 2034 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2035 struct sk_buff *skb, 2036 struct udp_offload *uoff); 2037 int (*gro_complete)(struct sk_buff *skb, 2038 int nhoff, 2039 struct udp_offload *uoff); 2040 }; 2041 2042 struct udp_offload { 2043 __be16 port; 2044 u8 ipproto; 2045 struct udp_offload_callbacks callbacks; 2046 }; 2047 2048 /* often modified stats are per cpu, other are shared (netdev->stats) */ 2049 struct pcpu_sw_netstats { 2050 u64 rx_packets; 2051 u64 rx_bytes; 2052 u64 tx_packets; 2053 u64 tx_bytes; 2054 struct u64_stats_sync syncp; 2055 }; 2056 2057 #define netdev_alloc_pcpu_stats(type) \ 2058 ({ \ 2059 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \ 2060 if (pcpu_stats) { \ 2061 int __cpu; \ 2062 for_each_possible_cpu(__cpu) { \ 2063 typeof(type) *stat; \ 2064 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2065 u64_stats_init(&stat->syncp); \ 2066 } \ 2067 } \ 2068 pcpu_stats; \ 2069 }) 2070 2071 #include <linux/notifier.h> 2072 2073 /* netdevice notifier chain. Please remember to update the rtnetlink 2074 * notification exclusion list in rtnetlink_event() when adding new 2075 * types. 2076 */ 2077 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2078 #define NETDEV_DOWN 0x0002 2079 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2080 detected a hardware crash and restarted 2081 - we can use this eg to kick tcp sessions 2082 once done */ 2083 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2084 #define NETDEV_REGISTER 0x0005 2085 #define NETDEV_UNREGISTER 0x0006 2086 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2087 #define NETDEV_CHANGEADDR 0x0008 2088 #define NETDEV_GOING_DOWN 0x0009 2089 #define NETDEV_CHANGENAME 0x000A 2090 #define NETDEV_FEAT_CHANGE 0x000B 2091 #define NETDEV_BONDING_FAILOVER 0x000C 2092 #define NETDEV_PRE_UP 0x000D 2093 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2094 #define NETDEV_POST_TYPE_CHANGE 0x000F 2095 #define NETDEV_POST_INIT 0x0010 2096 #define NETDEV_UNREGISTER_FINAL 0x0011 2097 #define NETDEV_RELEASE 0x0012 2098 #define NETDEV_NOTIFY_PEERS 0x0013 2099 #define NETDEV_JOIN 0x0014 2100 #define NETDEV_CHANGEUPPER 0x0015 2101 #define NETDEV_RESEND_IGMP 0x0016 2102 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2103 #define NETDEV_CHANGEINFODATA 0x0018 2104 #define NETDEV_BONDING_INFO 0x0019 2105 2106 int register_netdevice_notifier(struct notifier_block *nb); 2107 int unregister_netdevice_notifier(struct notifier_block *nb); 2108 2109 struct netdev_notifier_info { 2110 struct net_device *dev; 2111 }; 2112 2113 struct netdev_notifier_change_info { 2114 struct netdev_notifier_info info; /* must be first */ 2115 unsigned int flags_changed; 2116 }; 2117 2118 struct netdev_notifier_changeupper_info { 2119 struct netdev_notifier_info info; /* must be first */ 2120 struct net_device *upper_dev; /* new upper dev */ 2121 bool master; /* is upper dev master */ 2122 bool linking; /* is the nofication for link or unlink */ 2123 }; 2124 2125 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2126 struct net_device *dev) 2127 { 2128 info->dev = dev; 2129 } 2130 2131 static inline struct net_device * 2132 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2133 { 2134 return info->dev; 2135 } 2136 2137 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2138 2139 2140 extern rwlock_t dev_base_lock; /* Device list lock */ 2141 2142 #define for_each_netdev(net, d) \ 2143 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2144 #define for_each_netdev_reverse(net, d) \ 2145 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2146 #define for_each_netdev_rcu(net, d) \ 2147 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2148 #define for_each_netdev_safe(net, d, n) \ 2149 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2150 #define for_each_netdev_continue(net, d) \ 2151 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2152 #define for_each_netdev_continue_rcu(net, d) \ 2153 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2154 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2155 for_each_netdev_rcu(&init_net, slave) \ 2156 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2157 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2158 2159 static inline struct net_device *next_net_device(struct net_device *dev) 2160 { 2161 struct list_head *lh; 2162 struct net *net; 2163 2164 net = dev_net(dev); 2165 lh = dev->dev_list.next; 2166 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2167 } 2168 2169 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2170 { 2171 struct list_head *lh; 2172 struct net *net; 2173 2174 net = dev_net(dev); 2175 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2176 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2177 } 2178 2179 static inline struct net_device *first_net_device(struct net *net) 2180 { 2181 return list_empty(&net->dev_base_head) ? NULL : 2182 net_device_entry(net->dev_base_head.next); 2183 } 2184 2185 static inline struct net_device *first_net_device_rcu(struct net *net) 2186 { 2187 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2188 2189 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2190 } 2191 2192 int netdev_boot_setup_check(struct net_device *dev); 2193 unsigned long netdev_boot_base(const char *prefix, int unit); 2194 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2195 const char *hwaddr); 2196 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2197 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2198 void dev_add_pack(struct packet_type *pt); 2199 void dev_remove_pack(struct packet_type *pt); 2200 void __dev_remove_pack(struct packet_type *pt); 2201 void dev_add_offload(struct packet_offload *po); 2202 void dev_remove_offload(struct packet_offload *po); 2203 2204 int dev_get_iflink(const struct net_device *dev); 2205 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2206 unsigned short mask); 2207 struct net_device *dev_get_by_name(struct net *net, const char *name); 2208 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2209 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2210 int dev_alloc_name(struct net_device *dev, const char *name); 2211 int dev_open(struct net_device *dev); 2212 int dev_close(struct net_device *dev); 2213 int dev_close_many(struct list_head *head, bool unlink); 2214 void dev_disable_lro(struct net_device *dev); 2215 int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb); 2216 int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb); 2217 static inline int dev_queue_xmit(struct sk_buff *skb) 2218 { 2219 return dev_queue_xmit_sk(skb->sk, skb); 2220 } 2221 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2222 int register_netdevice(struct net_device *dev); 2223 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2224 void unregister_netdevice_many(struct list_head *head); 2225 static inline void unregister_netdevice(struct net_device *dev) 2226 { 2227 unregister_netdevice_queue(dev, NULL); 2228 } 2229 2230 int netdev_refcnt_read(const struct net_device *dev); 2231 void free_netdev(struct net_device *dev); 2232 void netdev_freemem(struct net_device *dev); 2233 void synchronize_net(void); 2234 int init_dummy_netdev(struct net_device *dev); 2235 2236 DECLARE_PER_CPU(int, xmit_recursion); 2237 static inline int dev_recursion_level(void) 2238 { 2239 return this_cpu_read(xmit_recursion); 2240 } 2241 2242 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2243 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2244 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2245 int netdev_get_name(struct net *net, char *name, int ifindex); 2246 int dev_restart(struct net_device *dev); 2247 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2248 2249 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2250 { 2251 return NAPI_GRO_CB(skb)->data_offset; 2252 } 2253 2254 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2255 { 2256 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2257 } 2258 2259 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2260 { 2261 NAPI_GRO_CB(skb)->data_offset += len; 2262 } 2263 2264 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2265 unsigned int offset) 2266 { 2267 return NAPI_GRO_CB(skb)->frag0 + offset; 2268 } 2269 2270 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2271 { 2272 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2273 } 2274 2275 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2276 unsigned int offset) 2277 { 2278 if (!pskb_may_pull(skb, hlen)) 2279 return NULL; 2280 2281 NAPI_GRO_CB(skb)->frag0 = NULL; 2282 NAPI_GRO_CB(skb)->frag0_len = 0; 2283 return skb->data + offset; 2284 } 2285 2286 static inline void *skb_gro_network_header(struct sk_buff *skb) 2287 { 2288 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2289 skb_network_offset(skb); 2290 } 2291 2292 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2293 const void *start, unsigned int len) 2294 { 2295 if (NAPI_GRO_CB(skb)->csum_valid) 2296 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2297 csum_partial(start, len, 0)); 2298 } 2299 2300 /* GRO checksum functions. These are logical equivalents of the normal 2301 * checksum functions (in skbuff.h) except that they operate on the GRO 2302 * offsets and fields in sk_buff. 2303 */ 2304 2305 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2306 2307 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2308 { 2309 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2310 } 2311 2312 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2313 bool zero_okay, 2314 __sum16 check) 2315 { 2316 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2317 skb_checksum_start_offset(skb) < 2318 skb_gro_offset(skb)) && 2319 !skb_at_gro_remcsum_start(skb) && 2320 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2321 (!zero_okay || check)); 2322 } 2323 2324 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2325 __wsum psum) 2326 { 2327 if (NAPI_GRO_CB(skb)->csum_valid && 2328 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2329 return 0; 2330 2331 NAPI_GRO_CB(skb)->csum = psum; 2332 2333 return __skb_gro_checksum_complete(skb); 2334 } 2335 2336 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2337 { 2338 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2339 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2340 NAPI_GRO_CB(skb)->csum_cnt--; 2341 } else { 2342 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2343 * verified a new top level checksum or an encapsulated one 2344 * during GRO. This saves work if we fallback to normal path. 2345 */ 2346 __skb_incr_checksum_unnecessary(skb); 2347 } 2348 } 2349 2350 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2351 compute_pseudo) \ 2352 ({ \ 2353 __sum16 __ret = 0; \ 2354 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2355 __ret = __skb_gro_checksum_validate_complete(skb, \ 2356 compute_pseudo(skb, proto)); \ 2357 if (__ret) \ 2358 __skb_mark_checksum_bad(skb); \ 2359 else \ 2360 skb_gro_incr_csum_unnecessary(skb); \ 2361 __ret; \ 2362 }) 2363 2364 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2365 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2366 2367 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2368 compute_pseudo) \ 2369 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2370 2371 #define skb_gro_checksum_simple_validate(skb) \ 2372 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2373 2374 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2375 { 2376 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2377 !NAPI_GRO_CB(skb)->csum_valid); 2378 } 2379 2380 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2381 __sum16 check, __wsum pseudo) 2382 { 2383 NAPI_GRO_CB(skb)->csum = ~pseudo; 2384 NAPI_GRO_CB(skb)->csum_valid = 1; 2385 } 2386 2387 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2388 do { \ 2389 if (__skb_gro_checksum_convert_check(skb)) \ 2390 __skb_gro_checksum_convert(skb, check, \ 2391 compute_pseudo(skb, proto)); \ 2392 } while (0) 2393 2394 struct gro_remcsum { 2395 int offset; 2396 __wsum delta; 2397 }; 2398 2399 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2400 { 2401 grc->offset = 0; 2402 grc->delta = 0; 2403 } 2404 2405 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2406 unsigned int off, size_t hdrlen, 2407 int start, int offset, 2408 struct gro_remcsum *grc, 2409 bool nopartial) 2410 { 2411 __wsum delta; 2412 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2413 2414 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2415 2416 if (!nopartial) { 2417 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2418 return ptr; 2419 } 2420 2421 ptr = skb_gro_header_fast(skb, off); 2422 if (skb_gro_header_hard(skb, off + plen)) { 2423 ptr = skb_gro_header_slow(skb, off + plen, off); 2424 if (!ptr) 2425 return NULL; 2426 } 2427 2428 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2429 start, offset); 2430 2431 /* Adjust skb->csum since we changed the packet */ 2432 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2433 2434 grc->offset = off + hdrlen + offset; 2435 grc->delta = delta; 2436 2437 return ptr; 2438 } 2439 2440 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2441 struct gro_remcsum *grc) 2442 { 2443 void *ptr; 2444 size_t plen = grc->offset + sizeof(u16); 2445 2446 if (!grc->delta) 2447 return; 2448 2449 ptr = skb_gro_header_fast(skb, grc->offset); 2450 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2451 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2452 if (!ptr) 2453 return; 2454 } 2455 2456 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2457 } 2458 2459 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2460 unsigned short type, 2461 const void *daddr, const void *saddr, 2462 unsigned int len) 2463 { 2464 if (!dev->header_ops || !dev->header_ops->create) 2465 return 0; 2466 2467 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2468 } 2469 2470 static inline int dev_parse_header(const struct sk_buff *skb, 2471 unsigned char *haddr) 2472 { 2473 const struct net_device *dev = skb->dev; 2474 2475 if (!dev->header_ops || !dev->header_ops->parse) 2476 return 0; 2477 return dev->header_ops->parse(skb, haddr); 2478 } 2479 2480 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2481 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2482 static inline int unregister_gifconf(unsigned int family) 2483 { 2484 return register_gifconf(family, NULL); 2485 } 2486 2487 #ifdef CONFIG_NET_FLOW_LIMIT 2488 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2489 struct sd_flow_limit { 2490 u64 count; 2491 unsigned int num_buckets; 2492 unsigned int history_head; 2493 u16 history[FLOW_LIMIT_HISTORY]; 2494 u8 buckets[]; 2495 }; 2496 2497 extern int netdev_flow_limit_table_len; 2498 #endif /* CONFIG_NET_FLOW_LIMIT */ 2499 2500 /* 2501 * Incoming packets are placed on per-cpu queues 2502 */ 2503 struct softnet_data { 2504 struct list_head poll_list; 2505 struct sk_buff_head process_queue; 2506 2507 /* stats */ 2508 unsigned int processed; 2509 unsigned int time_squeeze; 2510 unsigned int cpu_collision; 2511 unsigned int received_rps; 2512 #ifdef CONFIG_RPS 2513 struct softnet_data *rps_ipi_list; 2514 #endif 2515 #ifdef CONFIG_NET_FLOW_LIMIT 2516 struct sd_flow_limit __rcu *flow_limit; 2517 #endif 2518 struct Qdisc *output_queue; 2519 struct Qdisc **output_queue_tailp; 2520 struct sk_buff *completion_queue; 2521 2522 #ifdef CONFIG_RPS 2523 /* Elements below can be accessed between CPUs for RPS */ 2524 struct call_single_data csd ____cacheline_aligned_in_smp; 2525 struct softnet_data *rps_ipi_next; 2526 unsigned int cpu; 2527 unsigned int input_queue_head; 2528 unsigned int input_queue_tail; 2529 #endif 2530 unsigned int dropped; 2531 struct sk_buff_head input_pkt_queue; 2532 struct napi_struct backlog; 2533 2534 }; 2535 2536 static inline void input_queue_head_incr(struct softnet_data *sd) 2537 { 2538 #ifdef CONFIG_RPS 2539 sd->input_queue_head++; 2540 #endif 2541 } 2542 2543 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2544 unsigned int *qtail) 2545 { 2546 #ifdef CONFIG_RPS 2547 *qtail = ++sd->input_queue_tail; 2548 #endif 2549 } 2550 2551 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2552 2553 void __netif_schedule(struct Qdisc *q); 2554 void netif_schedule_queue(struct netdev_queue *txq); 2555 2556 static inline void netif_tx_schedule_all(struct net_device *dev) 2557 { 2558 unsigned int i; 2559 2560 for (i = 0; i < dev->num_tx_queues; i++) 2561 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2562 } 2563 2564 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2565 { 2566 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2567 } 2568 2569 /** 2570 * netif_start_queue - allow transmit 2571 * @dev: network device 2572 * 2573 * Allow upper layers to call the device hard_start_xmit routine. 2574 */ 2575 static inline void netif_start_queue(struct net_device *dev) 2576 { 2577 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2578 } 2579 2580 static inline void netif_tx_start_all_queues(struct net_device *dev) 2581 { 2582 unsigned int i; 2583 2584 for (i = 0; i < dev->num_tx_queues; i++) { 2585 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2586 netif_tx_start_queue(txq); 2587 } 2588 } 2589 2590 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2591 2592 /** 2593 * netif_wake_queue - restart transmit 2594 * @dev: network device 2595 * 2596 * Allow upper layers to call the device hard_start_xmit routine. 2597 * Used for flow control when transmit resources are available. 2598 */ 2599 static inline void netif_wake_queue(struct net_device *dev) 2600 { 2601 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2602 } 2603 2604 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2605 { 2606 unsigned int i; 2607 2608 for (i = 0; i < dev->num_tx_queues; i++) { 2609 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2610 netif_tx_wake_queue(txq); 2611 } 2612 } 2613 2614 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2615 { 2616 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2617 } 2618 2619 /** 2620 * netif_stop_queue - stop transmitted packets 2621 * @dev: network device 2622 * 2623 * Stop upper layers calling the device hard_start_xmit routine. 2624 * Used for flow control when transmit resources are unavailable. 2625 */ 2626 static inline void netif_stop_queue(struct net_device *dev) 2627 { 2628 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2629 } 2630 2631 void netif_tx_stop_all_queues(struct net_device *dev); 2632 2633 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2634 { 2635 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2636 } 2637 2638 /** 2639 * netif_queue_stopped - test if transmit queue is flowblocked 2640 * @dev: network device 2641 * 2642 * Test if transmit queue on device is currently unable to send. 2643 */ 2644 static inline bool netif_queue_stopped(const struct net_device *dev) 2645 { 2646 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2647 } 2648 2649 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2650 { 2651 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2652 } 2653 2654 static inline bool 2655 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2656 { 2657 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2658 } 2659 2660 static inline bool 2661 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2662 { 2663 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2664 } 2665 2666 /** 2667 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2668 * @dev_queue: pointer to transmit queue 2669 * 2670 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2671 * to give appropriate hint to the cpu. 2672 */ 2673 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2674 { 2675 #ifdef CONFIG_BQL 2676 prefetchw(&dev_queue->dql.num_queued); 2677 #endif 2678 } 2679 2680 /** 2681 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2682 * @dev_queue: pointer to transmit queue 2683 * 2684 * BQL enabled drivers might use this helper in their TX completion path, 2685 * to give appropriate hint to the cpu. 2686 */ 2687 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2688 { 2689 #ifdef CONFIG_BQL 2690 prefetchw(&dev_queue->dql.limit); 2691 #endif 2692 } 2693 2694 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2695 unsigned int bytes) 2696 { 2697 #ifdef CONFIG_BQL 2698 dql_queued(&dev_queue->dql, bytes); 2699 2700 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2701 return; 2702 2703 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2704 2705 /* 2706 * The XOFF flag must be set before checking the dql_avail below, 2707 * because in netdev_tx_completed_queue we update the dql_completed 2708 * before checking the XOFF flag. 2709 */ 2710 smp_mb(); 2711 2712 /* check again in case another CPU has just made room avail */ 2713 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2714 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2715 #endif 2716 } 2717 2718 /** 2719 * netdev_sent_queue - report the number of bytes queued to hardware 2720 * @dev: network device 2721 * @bytes: number of bytes queued to the hardware device queue 2722 * 2723 * Report the number of bytes queued for sending/completion to the network 2724 * device hardware queue. @bytes should be a good approximation and should 2725 * exactly match netdev_completed_queue() @bytes 2726 */ 2727 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2728 { 2729 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2730 } 2731 2732 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2733 unsigned int pkts, unsigned int bytes) 2734 { 2735 #ifdef CONFIG_BQL 2736 if (unlikely(!bytes)) 2737 return; 2738 2739 dql_completed(&dev_queue->dql, bytes); 2740 2741 /* 2742 * Without the memory barrier there is a small possiblity that 2743 * netdev_tx_sent_queue will miss the update and cause the queue to 2744 * be stopped forever 2745 */ 2746 smp_mb(); 2747 2748 if (dql_avail(&dev_queue->dql) < 0) 2749 return; 2750 2751 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2752 netif_schedule_queue(dev_queue); 2753 #endif 2754 } 2755 2756 /** 2757 * netdev_completed_queue - report bytes and packets completed by device 2758 * @dev: network device 2759 * @pkts: actual number of packets sent over the medium 2760 * @bytes: actual number of bytes sent over the medium 2761 * 2762 * Report the number of bytes and packets transmitted by the network device 2763 * hardware queue over the physical medium, @bytes must exactly match the 2764 * @bytes amount passed to netdev_sent_queue() 2765 */ 2766 static inline void netdev_completed_queue(struct net_device *dev, 2767 unsigned int pkts, unsigned int bytes) 2768 { 2769 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 2770 } 2771 2772 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 2773 { 2774 #ifdef CONFIG_BQL 2775 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 2776 dql_reset(&q->dql); 2777 #endif 2778 } 2779 2780 /** 2781 * netdev_reset_queue - reset the packets and bytes count of a network device 2782 * @dev_queue: network device 2783 * 2784 * Reset the bytes and packet count of a network device and clear the 2785 * software flow control OFF bit for this network device 2786 */ 2787 static inline void netdev_reset_queue(struct net_device *dev_queue) 2788 { 2789 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 2790 } 2791 2792 /** 2793 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 2794 * @dev: network device 2795 * @queue_index: given tx queue index 2796 * 2797 * Returns 0 if given tx queue index >= number of device tx queues, 2798 * otherwise returns the originally passed tx queue index. 2799 */ 2800 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 2801 { 2802 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2803 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 2804 dev->name, queue_index, 2805 dev->real_num_tx_queues); 2806 return 0; 2807 } 2808 2809 return queue_index; 2810 } 2811 2812 /** 2813 * netif_running - test if up 2814 * @dev: network device 2815 * 2816 * Test if the device has been brought up. 2817 */ 2818 static inline bool netif_running(const struct net_device *dev) 2819 { 2820 return test_bit(__LINK_STATE_START, &dev->state); 2821 } 2822 2823 /* 2824 * Routines to manage the subqueues on a device. We only need start 2825 * stop, and a check if it's stopped. All other device management is 2826 * done at the overall netdevice level. 2827 * Also test the device if we're multiqueue. 2828 */ 2829 2830 /** 2831 * netif_start_subqueue - allow sending packets on subqueue 2832 * @dev: network device 2833 * @queue_index: sub queue index 2834 * 2835 * Start individual transmit queue of a device with multiple transmit queues. 2836 */ 2837 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2838 { 2839 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2840 2841 netif_tx_start_queue(txq); 2842 } 2843 2844 /** 2845 * netif_stop_subqueue - stop sending packets on subqueue 2846 * @dev: network device 2847 * @queue_index: sub queue index 2848 * 2849 * Stop individual transmit queue of a device with multiple transmit queues. 2850 */ 2851 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2852 { 2853 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2854 netif_tx_stop_queue(txq); 2855 } 2856 2857 /** 2858 * netif_subqueue_stopped - test status of subqueue 2859 * @dev: network device 2860 * @queue_index: sub queue index 2861 * 2862 * Check individual transmit queue of a device with multiple transmit queues. 2863 */ 2864 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2865 u16 queue_index) 2866 { 2867 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2868 2869 return netif_tx_queue_stopped(txq); 2870 } 2871 2872 static inline bool netif_subqueue_stopped(const struct net_device *dev, 2873 struct sk_buff *skb) 2874 { 2875 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2876 } 2877 2878 void netif_wake_subqueue(struct net_device *dev, u16 queue_index); 2879 2880 #ifdef CONFIG_XPS 2881 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 2882 u16 index); 2883 #else 2884 static inline int netif_set_xps_queue(struct net_device *dev, 2885 const struct cpumask *mask, 2886 u16 index) 2887 { 2888 return 0; 2889 } 2890 #endif 2891 2892 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 2893 unsigned int num_tx_queues); 2894 2895 /* 2896 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2897 * as a distribution range limit for the returned value. 2898 */ 2899 static inline u16 skb_tx_hash(const struct net_device *dev, 2900 struct sk_buff *skb) 2901 { 2902 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2903 } 2904 2905 /** 2906 * netif_is_multiqueue - test if device has multiple transmit queues 2907 * @dev: network device 2908 * 2909 * Check if device has multiple transmit queues 2910 */ 2911 static inline bool netif_is_multiqueue(const struct net_device *dev) 2912 { 2913 return dev->num_tx_queues > 1; 2914 } 2915 2916 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 2917 2918 #ifdef CONFIG_SYSFS 2919 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 2920 #else 2921 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2922 unsigned int rxq) 2923 { 2924 return 0; 2925 } 2926 #endif 2927 2928 #ifdef CONFIG_SYSFS 2929 static inline unsigned int get_netdev_rx_queue_index( 2930 struct netdev_rx_queue *queue) 2931 { 2932 struct net_device *dev = queue->dev; 2933 int index = queue - dev->_rx; 2934 2935 BUG_ON(index >= dev->num_rx_queues); 2936 return index; 2937 } 2938 #endif 2939 2940 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2941 int netif_get_num_default_rss_queues(void); 2942 2943 enum skb_free_reason { 2944 SKB_REASON_CONSUMED, 2945 SKB_REASON_DROPPED, 2946 }; 2947 2948 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 2949 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 2950 2951 /* 2952 * It is not allowed to call kfree_skb() or consume_skb() from hardware 2953 * interrupt context or with hardware interrupts being disabled. 2954 * (in_irq() || irqs_disabled()) 2955 * 2956 * We provide four helpers that can be used in following contexts : 2957 * 2958 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 2959 * replacing kfree_skb(skb) 2960 * 2961 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 2962 * Typically used in place of consume_skb(skb) in TX completion path 2963 * 2964 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 2965 * replacing kfree_skb(skb) 2966 * 2967 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 2968 * and consumed a packet. Used in place of consume_skb(skb) 2969 */ 2970 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 2971 { 2972 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 2973 } 2974 2975 static inline void dev_consume_skb_irq(struct sk_buff *skb) 2976 { 2977 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 2978 } 2979 2980 static inline void dev_kfree_skb_any(struct sk_buff *skb) 2981 { 2982 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 2983 } 2984 2985 static inline void dev_consume_skb_any(struct sk_buff *skb) 2986 { 2987 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 2988 } 2989 2990 int netif_rx(struct sk_buff *skb); 2991 int netif_rx_ni(struct sk_buff *skb); 2992 int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb); 2993 static inline int netif_receive_skb(struct sk_buff *skb) 2994 { 2995 return netif_receive_skb_sk(skb->sk, skb); 2996 } 2997 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 2998 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 2999 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3000 gro_result_t napi_gro_frags(struct napi_struct *napi); 3001 struct packet_offload *gro_find_receive_by_type(__be16 type); 3002 struct packet_offload *gro_find_complete_by_type(__be16 type); 3003 3004 static inline void napi_free_frags(struct napi_struct *napi) 3005 { 3006 kfree_skb(napi->skb); 3007 napi->skb = NULL; 3008 } 3009 3010 int netdev_rx_handler_register(struct net_device *dev, 3011 rx_handler_func_t *rx_handler, 3012 void *rx_handler_data); 3013 void netdev_rx_handler_unregister(struct net_device *dev); 3014 3015 bool dev_valid_name(const char *name); 3016 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3017 int dev_ethtool(struct net *net, struct ifreq *); 3018 unsigned int dev_get_flags(const struct net_device *); 3019 int __dev_change_flags(struct net_device *, unsigned int flags); 3020 int dev_change_flags(struct net_device *, unsigned int); 3021 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3022 unsigned int gchanges); 3023 int dev_change_name(struct net_device *, const char *); 3024 int dev_set_alias(struct net_device *, const char *, size_t); 3025 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3026 int dev_set_mtu(struct net_device *, int); 3027 void dev_set_group(struct net_device *, int); 3028 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3029 int dev_change_carrier(struct net_device *, bool new_carrier); 3030 int dev_get_phys_port_id(struct net_device *dev, 3031 struct netdev_phys_item_id *ppid); 3032 int dev_get_phys_port_name(struct net_device *dev, 3033 char *name, size_t len); 3034 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3035 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3036 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3037 struct netdev_queue *txq, int *ret); 3038 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3039 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3040 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); 3041 3042 extern int netdev_budget; 3043 3044 /* Called by rtnetlink.c:rtnl_unlock() */ 3045 void netdev_run_todo(void); 3046 3047 /** 3048 * dev_put - release reference to device 3049 * @dev: network device 3050 * 3051 * Release reference to device to allow it to be freed. 3052 */ 3053 static inline void dev_put(struct net_device *dev) 3054 { 3055 this_cpu_dec(*dev->pcpu_refcnt); 3056 } 3057 3058 /** 3059 * dev_hold - get reference to device 3060 * @dev: network device 3061 * 3062 * Hold reference to device to keep it from being freed. 3063 */ 3064 static inline void dev_hold(struct net_device *dev) 3065 { 3066 this_cpu_inc(*dev->pcpu_refcnt); 3067 } 3068 3069 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3070 * and _off may be called from IRQ context, but it is caller 3071 * who is responsible for serialization of these calls. 3072 * 3073 * The name carrier is inappropriate, these functions should really be 3074 * called netif_lowerlayer_*() because they represent the state of any 3075 * kind of lower layer not just hardware media. 3076 */ 3077 3078 void linkwatch_init_dev(struct net_device *dev); 3079 void linkwatch_fire_event(struct net_device *dev); 3080 void linkwatch_forget_dev(struct net_device *dev); 3081 3082 /** 3083 * netif_carrier_ok - test if carrier present 3084 * @dev: network device 3085 * 3086 * Check if carrier is present on device 3087 */ 3088 static inline bool netif_carrier_ok(const struct net_device *dev) 3089 { 3090 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3091 } 3092 3093 unsigned long dev_trans_start(struct net_device *dev); 3094 3095 void __netdev_watchdog_up(struct net_device *dev); 3096 3097 void netif_carrier_on(struct net_device *dev); 3098 3099 void netif_carrier_off(struct net_device *dev); 3100 3101 /** 3102 * netif_dormant_on - mark device as dormant. 3103 * @dev: network device 3104 * 3105 * Mark device as dormant (as per RFC2863). 3106 * 3107 * The dormant state indicates that the relevant interface is not 3108 * actually in a condition to pass packets (i.e., it is not 'up') but is 3109 * in a "pending" state, waiting for some external event. For "on- 3110 * demand" interfaces, this new state identifies the situation where the 3111 * interface is waiting for events to place it in the up state. 3112 * 3113 */ 3114 static inline void netif_dormant_on(struct net_device *dev) 3115 { 3116 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3117 linkwatch_fire_event(dev); 3118 } 3119 3120 /** 3121 * netif_dormant_off - set device as not dormant. 3122 * @dev: network device 3123 * 3124 * Device is not in dormant state. 3125 */ 3126 static inline void netif_dormant_off(struct net_device *dev) 3127 { 3128 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3129 linkwatch_fire_event(dev); 3130 } 3131 3132 /** 3133 * netif_dormant - test if carrier present 3134 * @dev: network device 3135 * 3136 * Check if carrier is present on device 3137 */ 3138 static inline bool netif_dormant(const struct net_device *dev) 3139 { 3140 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3141 } 3142 3143 3144 /** 3145 * netif_oper_up - test if device is operational 3146 * @dev: network device 3147 * 3148 * Check if carrier is operational 3149 */ 3150 static inline bool netif_oper_up(const struct net_device *dev) 3151 { 3152 return (dev->operstate == IF_OPER_UP || 3153 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3154 } 3155 3156 /** 3157 * netif_device_present - is device available or removed 3158 * @dev: network device 3159 * 3160 * Check if device has not been removed from system. 3161 */ 3162 static inline bool netif_device_present(struct net_device *dev) 3163 { 3164 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3165 } 3166 3167 void netif_device_detach(struct net_device *dev); 3168 3169 void netif_device_attach(struct net_device *dev); 3170 3171 /* 3172 * Network interface message level settings 3173 */ 3174 3175 enum { 3176 NETIF_MSG_DRV = 0x0001, 3177 NETIF_MSG_PROBE = 0x0002, 3178 NETIF_MSG_LINK = 0x0004, 3179 NETIF_MSG_TIMER = 0x0008, 3180 NETIF_MSG_IFDOWN = 0x0010, 3181 NETIF_MSG_IFUP = 0x0020, 3182 NETIF_MSG_RX_ERR = 0x0040, 3183 NETIF_MSG_TX_ERR = 0x0080, 3184 NETIF_MSG_TX_QUEUED = 0x0100, 3185 NETIF_MSG_INTR = 0x0200, 3186 NETIF_MSG_TX_DONE = 0x0400, 3187 NETIF_MSG_RX_STATUS = 0x0800, 3188 NETIF_MSG_PKTDATA = 0x1000, 3189 NETIF_MSG_HW = 0x2000, 3190 NETIF_MSG_WOL = 0x4000, 3191 }; 3192 3193 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3194 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3195 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3196 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3197 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3198 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3199 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3200 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3201 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3202 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3203 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3204 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3205 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3206 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3207 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3208 3209 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3210 { 3211 /* use default */ 3212 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3213 return default_msg_enable_bits; 3214 if (debug_value == 0) /* no output */ 3215 return 0; 3216 /* set low N bits */ 3217 return (1 << debug_value) - 1; 3218 } 3219 3220 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3221 { 3222 spin_lock(&txq->_xmit_lock); 3223 txq->xmit_lock_owner = cpu; 3224 } 3225 3226 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3227 { 3228 spin_lock_bh(&txq->_xmit_lock); 3229 txq->xmit_lock_owner = smp_processor_id(); 3230 } 3231 3232 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3233 { 3234 bool ok = spin_trylock(&txq->_xmit_lock); 3235 if (likely(ok)) 3236 txq->xmit_lock_owner = smp_processor_id(); 3237 return ok; 3238 } 3239 3240 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3241 { 3242 txq->xmit_lock_owner = -1; 3243 spin_unlock(&txq->_xmit_lock); 3244 } 3245 3246 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3247 { 3248 txq->xmit_lock_owner = -1; 3249 spin_unlock_bh(&txq->_xmit_lock); 3250 } 3251 3252 static inline void txq_trans_update(struct netdev_queue *txq) 3253 { 3254 if (txq->xmit_lock_owner != -1) 3255 txq->trans_start = jiffies; 3256 } 3257 3258 /** 3259 * netif_tx_lock - grab network device transmit lock 3260 * @dev: network device 3261 * 3262 * Get network device transmit lock 3263 */ 3264 static inline void netif_tx_lock(struct net_device *dev) 3265 { 3266 unsigned int i; 3267 int cpu; 3268 3269 spin_lock(&dev->tx_global_lock); 3270 cpu = smp_processor_id(); 3271 for (i = 0; i < dev->num_tx_queues; i++) { 3272 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3273 3274 /* We are the only thread of execution doing a 3275 * freeze, but we have to grab the _xmit_lock in 3276 * order to synchronize with threads which are in 3277 * the ->hard_start_xmit() handler and already 3278 * checked the frozen bit. 3279 */ 3280 __netif_tx_lock(txq, cpu); 3281 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3282 __netif_tx_unlock(txq); 3283 } 3284 } 3285 3286 static inline void netif_tx_lock_bh(struct net_device *dev) 3287 { 3288 local_bh_disable(); 3289 netif_tx_lock(dev); 3290 } 3291 3292 static inline void netif_tx_unlock(struct net_device *dev) 3293 { 3294 unsigned int i; 3295 3296 for (i = 0; i < dev->num_tx_queues; i++) { 3297 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3298 3299 /* No need to grab the _xmit_lock here. If the 3300 * queue is not stopped for another reason, we 3301 * force a schedule. 3302 */ 3303 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3304 netif_schedule_queue(txq); 3305 } 3306 spin_unlock(&dev->tx_global_lock); 3307 } 3308 3309 static inline void netif_tx_unlock_bh(struct net_device *dev) 3310 { 3311 netif_tx_unlock(dev); 3312 local_bh_enable(); 3313 } 3314 3315 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3316 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3317 __netif_tx_lock(txq, cpu); \ 3318 } \ 3319 } 3320 3321 #define HARD_TX_TRYLOCK(dev, txq) \ 3322 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3323 __netif_tx_trylock(txq) : \ 3324 true ) 3325 3326 #define HARD_TX_UNLOCK(dev, txq) { \ 3327 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3328 __netif_tx_unlock(txq); \ 3329 } \ 3330 } 3331 3332 static inline void netif_tx_disable(struct net_device *dev) 3333 { 3334 unsigned int i; 3335 int cpu; 3336 3337 local_bh_disable(); 3338 cpu = smp_processor_id(); 3339 for (i = 0; i < dev->num_tx_queues; i++) { 3340 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3341 3342 __netif_tx_lock(txq, cpu); 3343 netif_tx_stop_queue(txq); 3344 __netif_tx_unlock(txq); 3345 } 3346 local_bh_enable(); 3347 } 3348 3349 static inline void netif_addr_lock(struct net_device *dev) 3350 { 3351 spin_lock(&dev->addr_list_lock); 3352 } 3353 3354 static inline void netif_addr_lock_nested(struct net_device *dev) 3355 { 3356 int subclass = SINGLE_DEPTH_NESTING; 3357 3358 if (dev->netdev_ops->ndo_get_lock_subclass) 3359 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3360 3361 spin_lock_nested(&dev->addr_list_lock, subclass); 3362 } 3363 3364 static inline void netif_addr_lock_bh(struct net_device *dev) 3365 { 3366 spin_lock_bh(&dev->addr_list_lock); 3367 } 3368 3369 static inline void netif_addr_unlock(struct net_device *dev) 3370 { 3371 spin_unlock(&dev->addr_list_lock); 3372 } 3373 3374 static inline void netif_addr_unlock_bh(struct net_device *dev) 3375 { 3376 spin_unlock_bh(&dev->addr_list_lock); 3377 } 3378 3379 /* 3380 * dev_addrs walker. Should be used only for read access. Call with 3381 * rcu_read_lock held. 3382 */ 3383 #define for_each_dev_addr(dev, ha) \ 3384 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3385 3386 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3387 3388 void ether_setup(struct net_device *dev); 3389 3390 /* Support for loadable net-drivers */ 3391 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3392 unsigned char name_assign_type, 3393 void (*setup)(struct net_device *), 3394 unsigned int txqs, unsigned int rxqs); 3395 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3396 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3397 3398 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3399 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3400 count) 3401 3402 int register_netdev(struct net_device *dev); 3403 void unregister_netdev(struct net_device *dev); 3404 3405 /* General hardware address lists handling functions */ 3406 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3407 struct netdev_hw_addr_list *from_list, int addr_len); 3408 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3409 struct netdev_hw_addr_list *from_list, int addr_len); 3410 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3411 struct net_device *dev, 3412 int (*sync)(struct net_device *, const unsigned char *), 3413 int (*unsync)(struct net_device *, 3414 const unsigned char *)); 3415 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3416 struct net_device *dev, 3417 int (*unsync)(struct net_device *, 3418 const unsigned char *)); 3419 void __hw_addr_init(struct netdev_hw_addr_list *list); 3420 3421 /* Functions used for device addresses handling */ 3422 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3423 unsigned char addr_type); 3424 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3425 unsigned char addr_type); 3426 void dev_addr_flush(struct net_device *dev); 3427 int dev_addr_init(struct net_device *dev); 3428 3429 /* Functions used for unicast addresses handling */ 3430 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3431 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3432 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3433 int dev_uc_sync(struct net_device *to, struct net_device *from); 3434 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3435 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3436 void dev_uc_flush(struct net_device *dev); 3437 void dev_uc_init(struct net_device *dev); 3438 3439 /** 3440 * __dev_uc_sync - Synchonize device's unicast list 3441 * @dev: device to sync 3442 * @sync: function to call if address should be added 3443 * @unsync: function to call if address should be removed 3444 * 3445 * Add newly added addresses to the interface, and release 3446 * addresses that have been deleted. 3447 **/ 3448 static inline int __dev_uc_sync(struct net_device *dev, 3449 int (*sync)(struct net_device *, 3450 const unsigned char *), 3451 int (*unsync)(struct net_device *, 3452 const unsigned char *)) 3453 { 3454 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3455 } 3456 3457 /** 3458 * __dev_uc_unsync - Remove synchronized addresses from device 3459 * @dev: device to sync 3460 * @unsync: function to call if address should be removed 3461 * 3462 * Remove all addresses that were added to the device by dev_uc_sync(). 3463 **/ 3464 static inline void __dev_uc_unsync(struct net_device *dev, 3465 int (*unsync)(struct net_device *, 3466 const unsigned char *)) 3467 { 3468 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3469 } 3470 3471 /* Functions used for multicast addresses handling */ 3472 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3473 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3474 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3475 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3476 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3477 int dev_mc_sync(struct net_device *to, struct net_device *from); 3478 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3479 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3480 void dev_mc_flush(struct net_device *dev); 3481 void dev_mc_init(struct net_device *dev); 3482 3483 /** 3484 * __dev_mc_sync - Synchonize device's multicast list 3485 * @dev: device to sync 3486 * @sync: function to call if address should be added 3487 * @unsync: function to call if address should be removed 3488 * 3489 * Add newly added addresses to the interface, and release 3490 * addresses that have been deleted. 3491 **/ 3492 static inline int __dev_mc_sync(struct net_device *dev, 3493 int (*sync)(struct net_device *, 3494 const unsigned char *), 3495 int (*unsync)(struct net_device *, 3496 const unsigned char *)) 3497 { 3498 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3499 } 3500 3501 /** 3502 * __dev_mc_unsync - Remove synchronized addresses from device 3503 * @dev: device to sync 3504 * @unsync: function to call if address should be removed 3505 * 3506 * Remove all addresses that were added to the device by dev_mc_sync(). 3507 **/ 3508 static inline void __dev_mc_unsync(struct net_device *dev, 3509 int (*unsync)(struct net_device *, 3510 const unsigned char *)) 3511 { 3512 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3513 } 3514 3515 /* Functions used for secondary unicast and multicast support */ 3516 void dev_set_rx_mode(struct net_device *dev); 3517 void __dev_set_rx_mode(struct net_device *dev); 3518 int dev_set_promiscuity(struct net_device *dev, int inc); 3519 int dev_set_allmulti(struct net_device *dev, int inc); 3520 void netdev_state_change(struct net_device *dev); 3521 void netdev_notify_peers(struct net_device *dev); 3522 void netdev_features_change(struct net_device *dev); 3523 /* Load a device via the kmod */ 3524 void dev_load(struct net *net, const char *name); 3525 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3526 struct rtnl_link_stats64 *storage); 3527 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3528 const struct net_device_stats *netdev_stats); 3529 3530 extern int netdev_max_backlog; 3531 extern int netdev_tstamp_prequeue; 3532 extern int weight_p; 3533 extern int bpf_jit_enable; 3534 3535 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3536 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3537 struct list_head **iter); 3538 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3539 struct list_head **iter); 3540 3541 /* iterate through upper list, must be called under RCU read lock */ 3542 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3543 for (iter = &(dev)->adj_list.upper, \ 3544 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3545 updev; \ 3546 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3547 3548 /* iterate through upper list, must be called under RCU read lock */ 3549 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ 3550 for (iter = &(dev)->all_adj_list.upper, \ 3551 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ 3552 updev; \ 3553 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) 3554 3555 void *netdev_lower_get_next_private(struct net_device *dev, 3556 struct list_head **iter); 3557 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3558 struct list_head **iter); 3559 3560 #define netdev_for_each_lower_private(dev, priv, iter) \ 3561 for (iter = (dev)->adj_list.lower.next, \ 3562 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3563 priv; \ 3564 priv = netdev_lower_get_next_private(dev, &(iter))) 3565 3566 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3567 for (iter = &(dev)->adj_list.lower, \ 3568 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3569 priv; \ 3570 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3571 3572 void *netdev_lower_get_next(struct net_device *dev, 3573 struct list_head **iter); 3574 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3575 for (iter = &(dev)->adj_list.lower, \ 3576 ldev = netdev_lower_get_next(dev, &(iter)); \ 3577 ldev; \ 3578 ldev = netdev_lower_get_next(dev, &(iter))) 3579 3580 void *netdev_adjacent_get_private(struct list_head *adj_list); 3581 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3582 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3583 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3584 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3585 int netdev_master_upper_dev_link(struct net_device *dev, 3586 struct net_device *upper_dev); 3587 int netdev_master_upper_dev_link_private(struct net_device *dev, 3588 struct net_device *upper_dev, 3589 void *private); 3590 void netdev_upper_dev_unlink(struct net_device *dev, 3591 struct net_device *upper_dev); 3592 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3593 void *netdev_lower_dev_get_private(struct net_device *dev, 3594 struct net_device *lower_dev); 3595 3596 /* RSS keys are 40 or 52 bytes long */ 3597 #define NETDEV_RSS_KEY_LEN 52 3598 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN]; 3599 void netdev_rss_key_fill(void *buffer, size_t len); 3600 3601 int dev_get_nest_level(struct net_device *dev, 3602 bool (*type_check)(struct net_device *dev)); 3603 int skb_checksum_help(struct sk_buff *skb); 3604 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3605 netdev_features_t features, bool tx_path); 3606 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3607 netdev_features_t features); 3608 3609 struct netdev_bonding_info { 3610 ifslave slave; 3611 ifbond master; 3612 }; 3613 3614 struct netdev_notifier_bonding_info { 3615 struct netdev_notifier_info info; /* must be first */ 3616 struct netdev_bonding_info bonding_info; 3617 }; 3618 3619 void netdev_bonding_info_change(struct net_device *dev, 3620 struct netdev_bonding_info *bonding_info); 3621 3622 static inline 3623 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3624 { 3625 return __skb_gso_segment(skb, features, true); 3626 } 3627 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3628 3629 static inline bool can_checksum_protocol(netdev_features_t features, 3630 __be16 protocol) 3631 { 3632 return ((features & NETIF_F_GEN_CSUM) || 3633 ((features & NETIF_F_V4_CSUM) && 3634 protocol == htons(ETH_P_IP)) || 3635 ((features & NETIF_F_V6_CSUM) && 3636 protocol == htons(ETH_P_IPV6)) || 3637 ((features & NETIF_F_FCOE_CRC) && 3638 protocol == htons(ETH_P_FCOE))); 3639 } 3640 3641 #ifdef CONFIG_BUG 3642 void netdev_rx_csum_fault(struct net_device *dev); 3643 #else 3644 static inline void netdev_rx_csum_fault(struct net_device *dev) 3645 { 3646 } 3647 #endif 3648 /* rx skb timestamps */ 3649 void net_enable_timestamp(void); 3650 void net_disable_timestamp(void); 3651 3652 #ifdef CONFIG_PROC_FS 3653 int __init dev_proc_init(void); 3654 #else 3655 #define dev_proc_init() 0 3656 #endif 3657 3658 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3659 struct sk_buff *skb, struct net_device *dev, 3660 bool more) 3661 { 3662 skb->xmit_more = more ? 1 : 0; 3663 return ops->ndo_start_xmit(skb, dev); 3664 } 3665 3666 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 3667 struct netdev_queue *txq, bool more) 3668 { 3669 const struct net_device_ops *ops = dev->netdev_ops; 3670 int rc; 3671 3672 rc = __netdev_start_xmit(ops, skb, dev, more); 3673 if (rc == NETDEV_TX_OK) 3674 txq_trans_update(txq); 3675 3676 return rc; 3677 } 3678 3679 int netdev_class_create_file_ns(struct class_attribute *class_attr, 3680 const void *ns); 3681 void netdev_class_remove_file_ns(struct class_attribute *class_attr, 3682 const void *ns); 3683 3684 static inline int netdev_class_create_file(struct class_attribute *class_attr) 3685 { 3686 return netdev_class_create_file_ns(class_attr, NULL); 3687 } 3688 3689 static inline void netdev_class_remove_file(struct class_attribute *class_attr) 3690 { 3691 netdev_class_remove_file_ns(class_attr, NULL); 3692 } 3693 3694 extern struct kobj_ns_type_operations net_ns_type_operations; 3695 3696 const char *netdev_drivername(const struct net_device *dev); 3697 3698 void linkwatch_run_queue(void); 3699 3700 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 3701 netdev_features_t f2) 3702 { 3703 if (f1 & NETIF_F_GEN_CSUM) 3704 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3705 if (f2 & NETIF_F_GEN_CSUM) 3706 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3707 f1 &= f2; 3708 if (f1 & NETIF_F_GEN_CSUM) 3709 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 3710 3711 return f1; 3712 } 3713 3714 static inline netdev_features_t netdev_get_wanted_features( 3715 struct net_device *dev) 3716 { 3717 return (dev->features & ~dev->hw_features) | dev->wanted_features; 3718 } 3719 netdev_features_t netdev_increment_features(netdev_features_t all, 3720 netdev_features_t one, netdev_features_t mask); 3721 3722 /* Allow TSO being used on stacked device : 3723 * Performing the GSO segmentation before last device 3724 * is a performance improvement. 3725 */ 3726 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 3727 netdev_features_t mask) 3728 { 3729 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 3730 } 3731 3732 int __netdev_update_features(struct net_device *dev); 3733 void netdev_update_features(struct net_device *dev); 3734 void netdev_change_features(struct net_device *dev); 3735 3736 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 3737 struct net_device *dev); 3738 3739 netdev_features_t passthru_features_check(struct sk_buff *skb, 3740 struct net_device *dev, 3741 netdev_features_t features); 3742 netdev_features_t netif_skb_features(struct sk_buff *skb); 3743 3744 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 3745 { 3746 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 3747 3748 /* check flags correspondence */ 3749 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 3750 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 3751 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 3752 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 3753 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 3754 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 3755 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 3756 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 3757 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT)); 3758 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT)); 3759 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 3760 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 3761 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 3762 3763 return (features & feature) == feature; 3764 } 3765 3766 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 3767 { 3768 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 3769 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 3770 } 3771 3772 static inline bool netif_needs_gso(struct sk_buff *skb, 3773 netdev_features_t features) 3774 { 3775 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 3776 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 3777 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 3778 } 3779 3780 static inline void netif_set_gso_max_size(struct net_device *dev, 3781 unsigned int size) 3782 { 3783 dev->gso_max_size = size; 3784 } 3785 3786 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 3787 int pulled_hlen, u16 mac_offset, 3788 int mac_len) 3789 { 3790 skb->protocol = protocol; 3791 skb->encapsulation = 1; 3792 skb_push(skb, pulled_hlen); 3793 skb_reset_transport_header(skb); 3794 skb->mac_header = mac_offset; 3795 skb->network_header = skb->mac_header + mac_len; 3796 skb->mac_len = mac_len; 3797 } 3798 3799 static inline bool netif_is_macvlan(struct net_device *dev) 3800 { 3801 return dev->priv_flags & IFF_MACVLAN; 3802 } 3803 3804 static inline bool netif_is_macvlan_port(struct net_device *dev) 3805 { 3806 return dev->priv_flags & IFF_MACVLAN_PORT; 3807 } 3808 3809 static inline bool netif_is_ipvlan(struct net_device *dev) 3810 { 3811 return dev->priv_flags & IFF_IPVLAN_SLAVE; 3812 } 3813 3814 static inline bool netif_is_ipvlan_port(struct net_device *dev) 3815 { 3816 return dev->priv_flags & IFF_IPVLAN_MASTER; 3817 } 3818 3819 static inline bool netif_is_bond_master(struct net_device *dev) 3820 { 3821 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 3822 } 3823 3824 static inline bool netif_is_bond_slave(struct net_device *dev) 3825 { 3826 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 3827 } 3828 3829 static inline bool netif_supports_nofcs(struct net_device *dev) 3830 { 3831 return dev->priv_flags & IFF_SUPP_NOFCS; 3832 } 3833 3834 static inline bool netif_is_vrf(const struct net_device *dev) 3835 { 3836 return dev->priv_flags & IFF_VRF_MASTER; 3837 } 3838 3839 static inline bool netif_is_bridge_master(const struct net_device *dev) 3840 { 3841 return dev->priv_flags & IFF_EBRIDGE; 3842 } 3843 3844 static inline bool netif_is_ovs_master(const struct net_device *dev) 3845 { 3846 return dev->priv_flags & IFF_OPENVSWITCH; 3847 } 3848 3849 static inline bool netif_index_is_vrf(struct net *net, int ifindex) 3850 { 3851 bool rc = false; 3852 3853 #if IS_ENABLED(CONFIG_NET_VRF) 3854 struct net_device *dev; 3855 3856 if (ifindex == 0) 3857 return false; 3858 3859 rcu_read_lock(); 3860 3861 dev = dev_get_by_index_rcu(net, ifindex); 3862 if (dev) 3863 rc = netif_is_vrf(dev); 3864 3865 rcu_read_unlock(); 3866 #endif 3867 return rc; 3868 } 3869 3870 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 3871 static inline void netif_keep_dst(struct net_device *dev) 3872 { 3873 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 3874 } 3875 3876 extern struct pernet_operations __net_initdata loopback_net_ops; 3877 3878 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 3879 3880 /* netdev_printk helpers, similar to dev_printk */ 3881 3882 static inline const char *netdev_name(const struct net_device *dev) 3883 { 3884 if (!dev->name[0] || strchr(dev->name, '%')) 3885 return "(unnamed net_device)"; 3886 return dev->name; 3887 } 3888 3889 static inline const char *netdev_reg_state(const struct net_device *dev) 3890 { 3891 switch (dev->reg_state) { 3892 case NETREG_UNINITIALIZED: return " (uninitialized)"; 3893 case NETREG_REGISTERED: return ""; 3894 case NETREG_UNREGISTERING: return " (unregistering)"; 3895 case NETREG_UNREGISTERED: return " (unregistered)"; 3896 case NETREG_RELEASED: return " (released)"; 3897 case NETREG_DUMMY: return " (dummy)"; 3898 } 3899 3900 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 3901 return " (unknown)"; 3902 } 3903 3904 __printf(3, 4) 3905 void netdev_printk(const char *level, const struct net_device *dev, 3906 const char *format, ...); 3907 __printf(2, 3) 3908 void netdev_emerg(const struct net_device *dev, const char *format, ...); 3909 __printf(2, 3) 3910 void netdev_alert(const struct net_device *dev, const char *format, ...); 3911 __printf(2, 3) 3912 void netdev_crit(const struct net_device *dev, const char *format, ...); 3913 __printf(2, 3) 3914 void netdev_err(const struct net_device *dev, const char *format, ...); 3915 __printf(2, 3) 3916 void netdev_warn(const struct net_device *dev, const char *format, ...); 3917 __printf(2, 3) 3918 void netdev_notice(const struct net_device *dev, const char *format, ...); 3919 __printf(2, 3) 3920 void netdev_info(const struct net_device *dev, const char *format, ...); 3921 3922 #define MODULE_ALIAS_NETDEV(device) \ 3923 MODULE_ALIAS("netdev-" device) 3924 3925 #if defined(CONFIG_DYNAMIC_DEBUG) 3926 #define netdev_dbg(__dev, format, args...) \ 3927 do { \ 3928 dynamic_netdev_dbg(__dev, format, ##args); \ 3929 } while (0) 3930 #elif defined(DEBUG) 3931 #define netdev_dbg(__dev, format, args...) \ 3932 netdev_printk(KERN_DEBUG, __dev, format, ##args) 3933 #else 3934 #define netdev_dbg(__dev, format, args...) \ 3935 ({ \ 3936 if (0) \ 3937 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 3938 }) 3939 #endif 3940 3941 #if defined(VERBOSE_DEBUG) 3942 #define netdev_vdbg netdev_dbg 3943 #else 3944 3945 #define netdev_vdbg(dev, format, args...) \ 3946 ({ \ 3947 if (0) \ 3948 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 3949 0; \ 3950 }) 3951 #endif 3952 3953 /* 3954 * netdev_WARN() acts like dev_printk(), but with the key difference 3955 * of using a WARN/WARN_ON to get the message out, including the 3956 * file/line information and a backtrace. 3957 */ 3958 #define netdev_WARN(dev, format, args...) \ 3959 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 3960 netdev_reg_state(dev), ##args) 3961 3962 /* netif printk helpers, similar to netdev_printk */ 3963 3964 #define netif_printk(priv, type, level, dev, fmt, args...) \ 3965 do { \ 3966 if (netif_msg_##type(priv)) \ 3967 netdev_printk(level, (dev), fmt, ##args); \ 3968 } while (0) 3969 3970 #define netif_level(level, priv, type, dev, fmt, args...) \ 3971 do { \ 3972 if (netif_msg_##type(priv)) \ 3973 netdev_##level(dev, fmt, ##args); \ 3974 } while (0) 3975 3976 #define netif_emerg(priv, type, dev, fmt, args...) \ 3977 netif_level(emerg, priv, type, dev, fmt, ##args) 3978 #define netif_alert(priv, type, dev, fmt, args...) \ 3979 netif_level(alert, priv, type, dev, fmt, ##args) 3980 #define netif_crit(priv, type, dev, fmt, args...) \ 3981 netif_level(crit, priv, type, dev, fmt, ##args) 3982 #define netif_err(priv, type, dev, fmt, args...) \ 3983 netif_level(err, priv, type, dev, fmt, ##args) 3984 #define netif_warn(priv, type, dev, fmt, args...) \ 3985 netif_level(warn, priv, type, dev, fmt, ##args) 3986 #define netif_notice(priv, type, dev, fmt, args...) \ 3987 netif_level(notice, priv, type, dev, fmt, ##args) 3988 #define netif_info(priv, type, dev, fmt, args...) \ 3989 netif_level(info, priv, type, dev, fmt, ##args) 3990 3991 #if defined(CONFIG_DYNAMIC_DEBUG) 3992 #define netif_dbg(priv, type, netdev, format, args...) \ 3993 do { \ 3994 if (netif_msg_##type(priv)) \ 3995 dynamic_netdev_dbg(netdev, format, ##args); \ 3996 } while (0) 3997 #elif defined(DEBUG) 3998 #define netif_dbg(priv, type, dev, format, args...) \ 3999 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4000 #else 4001 #define netif_dbg(priv, type, dev, format, args...) \ 4002 ({ \ 4003 if (0) \ 4004 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4005 0; \ 4006 }) 4007 #endif 4008 4009 #if defined(VERBOSE_DEBUG) 4010 #define netif_vdbg netif_dbg 4011 #else 4012 #define netif_vdbg(priv, type, dev, format, args...) \ 4013 ({ \ 4014 if (0) \ 4015 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4016 0; \ 4017 }) 4018 #endif 4019 4020 /* 4021 * The list of packet types we will receive (as opposed to discard) 4022 * and the routines to invoke. 4023 * 4024 * Why 16. Because with 16 the only overlap we get on a hash of the 4025 * low nibble of the protocol value is RARP/SNAP/X.25. 4026 * 4027 * NOTE: That is no longer true with the addition of VLAN tags. Not 4028 * sure which should go first, but I bet it won't make much 4029 * difference if we are running VLANs. The good news is that 4030 * this protocol won't be in the list unless compiled in, so 4031 * the average user (w/out VLANs) will not be adversely affected. 4032 * --BLG 4033 * 4034 * 0800 IP 4035 * 8100 802.1Q VLAN 4036 * 0001 802.3 4037 * 0002 AX.25 4038 * 0004 802.2 4039 * 8035 RARP 4040 * 0005 SNAP 4041 * 0805 X.25 4042 * 0806 ARP 4043 * 8137 IPX 4044 * 0009 Localtalk 4045 * 86DD IPv6 4046 */ 4047 #define PTYPE_HASH_SIZE (16) 4048 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4049 4050 #endif /* _LINUX_NETDEVICE_H */ 4051